* Add first version of streaming JPEG decoder.

Benchmark before:
```
Encoding                  kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------------
jpeg:q90                    13270  4838710    2.9169834  63.705 161.410   2.40070152   0.68663652  2.002907346396      0
jpeg:q90:djxl8              13270  4838710    2.9169834  63.213 112.574   2.17603183   0.62092407  1.811225220808      0
jpeg:q90:djxl16             13270  4838710    2.9169834  63.008 109.120   2.16807723   0.61567343  1.795909190468      0
jpeg:yuv420:q90             13270  3660915    2.2069577  93.729 238.752   6.07282543   0.95039908  2.097490592679      0
jpeg:yuv420:q90:djxl8       13270  3660915    2.2069577  93.761 156.263   6.05725956   0.91959307  2.029503031911      0
jpeg:yuv420:q90:djxl16      13270  3660915    2.2069577  95.847 152.478   6.05378914   0.91638444  2.022421724515      0
jpeg:libjxl                 13270  3197015    1.9272988  18.429  99.632   2.19571280   0.70762796  1.363810502985      0
jpeg:libjxl:djxl8           13270  3197015    1.9272988  18.652  86.314   2.25490046   0.70510193  1.358942071174      0
jpeg:libjxl:djxl16          13270  3197015    1.9272988  18.639  84.488   2.29413509   0.69006634  1.329964006217      0
```

Benchmark after:
```
Encoding                  kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------------
jpeg:q90                    13270  4838710    2.9169834  64.733 166.560   2.40070152   0.68663652  2.002907346396      0
jpeg:q90:djxl8              13270  4838710    2.9169834  63.666 105.345   2.17603183   0.62092407  1.811225220808      0
jpeg:q90:djxl16             13270  4838710    2.9169834  63.061 104.653   2.16807723   0.61567343  1.795909190468      0
jpeg:yuv420:q90             13270  3660915    2.2069577  96.172 246.808   6.07282543   0.95039908  2.097490592679      0
jpeg:yuv420:q90:djxl8       13270  3660915    2.2069577  97.151 156.914   6.05725956   0.91959307  2.029503031911      0
jpeg:yuv420:q90:djxl16      13270  3660915    2.2069577  97.868 153.058   6.05378914   0.91638444  2.022421724515      0
jpeg:libjxl                 13270  3197015    1.9272988  18.323  97.172   2.19571280   0.70762796  1.363810502985      0
jpeg:libjxl:djxl8           13270  3197015    1.9272988  18.660  76.811   2.25490046   0.70510193  1.358942071174      0
jpeg:libjxl:djxl16          13270  3197015    1.9272988  18.711  76.222   2.29413509   0.69006634  1.329964006217      0
```

* Dont save MCU coding state if there are enough bytes in buffer.

Benchmark before:
```
Encoding                 kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
-----------------------------------------------------------------------------------------------------------------------
jpeg:q90:djxl8             13270  4838710    2.9169834  65.817 112.747   2.17603183   0.62092407  1.811225220808      0
jpeg:yuv420:q90:djxl8      13270  3660915    2.2069577  96.322 164.534   6.05725956   0.91959307  2.029503031911      0
jpeg:libjxl:djxl8          13270  3197015    1.9272988  19.046  79.233   2.25490046   0.70510193  1.358942071174      0
```

Benchmark after:
```
Encoding                 kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
-----------------------------------------------------------------------------------------------------------------------
jpeg:q90:djxl8             13270  4838710    2.9169834  66.330 118.045   2.17603183   0.62092407  1.811225220808      0
jpeg:yuv420:q90:djxl8      13270  3660915    2.2069577  96.814 169.467   6.05725956   0.91959307  2.029503031911      0
jpeg:libjxl:djxl8          13270  3197015    1.9272988  18.882  82.873   2.25490046   0.70510193  1.358942071174      0
```

* Fix JPEG reconstruction error.

Fixes libjxl#1790.

* tools/libjxl_test.c: avoid function declaration without prototype (libjxl#1867)

Avoid declaring int main() without a prototype as this is deprecated
in all versions of C.

* Make jpeg decoder more self-contained. (libjxl#1866)

Add specialized 8x8 IDCT and transpose from jpeg xl.

* Encode intrinsic dimensions (libjxl#1860)

Make the encoder store the `intrinsic_xsize` and `intrinsic_ysize` from JxlBasicInfo. Setting these previously had no effect.

* Add a build mode without box decoding and jpeg transcoding.

In this case brotli decoder is also not needed.

Brotli compressed WASM binary size is 155k.

* XYB-jpeg: jpeg-specific clustering and more progression. (libjxl#1875)

Benchmark before:
```
Encoding            kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------
jpeg:libjxl:d1.0      13270  3197015    1.9272988   5.469  29.993   2.19093180   0.70477101  1.358304300765      0
```

Benchmark after:
```
Encoding            kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------
jpeg:libjxl:d1.0      13270  3163678    1.9072018   4.031  25.939   2.19093180   0.70477101  1.344140529099      0
```

* fast_lossless: rename FastLosslessEncode

Adding a "Jxl" prefix disambiguates it when combining it with e.g. a
fast PNG encoder or fast WebP encoder. Some lines were subsequently
re-formatted to fit within 80 columns.

Also wrap an "extern C" around that function.

Also wrap an anonymous namespace around the implementation, so that the
fjxl BitWriter doesn't conflict with e.g. fpnge's BitWriter.

Also fix some -Wunknown-pragmas and -Wsign-compare g++ warnings.

Fixes libjxl#1855

* lint and author file

* Simplex-search optimized xyb-jpeg quantization matrix parameters. (libjxl#1877)

Benchmark before:
```
Encoding            kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------
jpeg:libjxl:d0.5      13270  4804698    2.8964795   3.723  22.242   1.56636882   0.47385977  1.372525138424      0
jpeg:libjxl:d0.8      13270  3636329    2.1921362   4.089  25.696   1.87652230   0.61543190  1.349110553678      0
jpeg:libjxl:d0.9      13270  3378595    2.0367630   3.790  23.042   2.17102480   0.66006343  1.344392804990      0
jpeg:libjxl:d1.0      13270  3163678    1.9072018   4.064  24.490   2.19093180   0.70477101  1.344140529099      0
jpeg:libjxl:d1.1      13270  2984772    1.7993495   3.966  26.363   2.11484838   0.75048089  1.350377429818      0
jpeg:libjxl:d1.2      13270  2828203    1.7049630   4.278  28.153   2.29096746   0.78898283  1.345186506645      0
jpeg:libjxl:d1.5      13270  2447451    1.4754292   4.083  26.287   3.02021217   0.91575531  1.351132125102      0
jpeg:libjxl:d2.0      13270  2015903    1.2152734   4.220  31.476   3.28036547   1.09683000  1.332948351442      0
Aggregate:            13270  3064488    1.8474056   4.023  25.829   2.25605616   0.73004342  1.348686280677      0
```

Benchmark after:
```
Encoding            kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------
jpeg:libjxl:d0.5      13270  4654354    2.8058457   3.971  20.967   1.54231071   0.46554951  1.306260075416      0
jpeg:libjxl:d0.8      13270  3482816    2.0995920   4.085  24.342   1.70935893   0.61929160  1.300259661216      0
jpeg:libjxl:d0.9      13270  3223473    1.9432488   4.163  22.163   1.84545839   0.67493955  1.311575456149      0
jpeg:libjxl:d1.0      13270  3028977    1.8259982   4.598  25.571   1.90735793   0.71123393  1.298711878895      0
jpeg:libjxl:d1.1      13270  2833331    1.7080543   4.518  26.407   2.19600940   0.76419471  1.305286098499      0
jpeg:libjxl:d1.2      13270  2694178    1.6241669   3.971  25.394   2.27468419   0.79893476  1.297603396949      0
jpeg:libjxl:d1.5      13270  2307256    1.3909136   4.331  26.717   2.60481334   0.94243889  1.310851074545      0
jpeg:libjxl:d2.0      13270  1894603    1.1421485   4.573  33.426   3.18599176   1.12675147  1.286917548072      0
Aggregate:            13270  2919735    1.7601424   4.269  25.400   2.10445561   0.73980436  1.302161048110      0
```

* don't compile debug print tree code (libjxl#1881)

* Read the CICP tag from ICC profiles if present

* Remove xyb-jpegs dependence on JXL's parameterized quant matrix.

Optimize the set of remaining 105 free AC quant weights (since we
want the quant matrix to be symmetric).

Benchmark before:
```
Encoding            kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------
jpeg:libjxl:d0.5      13270  4654354    2.8058457   3.663  20.463   1.54231071   0.46554951  1.306260075416      0
jpeg:libjxl:d0.8      13270  3482816    2.0995920   3.796  25.002   1.70935893   0.61929160  1.300259661216      0
jpeg:libjxl:d0.9      13270  3223473    1.9432488   3.822  22.895   1.84545839   0.67493955  1.311575456149      0
jpeg:libjxl:d1.0      13270  3028977    1.8259982   3.726  21.924   1.90735793   0.71123393  1.298711878895      0
jpeg:libjxl:d1.1      13270  2833331    1.7080543   3.716  22.769   2.19600940   0.76419471  1.305286098499      0
jpeg:libjxl:d1.2      13270  2694178    1.6241669   3.751  23.004   2.27468419   0.79893476  1.297603396949      0
jpeg:libjxl:d1.5      13270  2307256    1.3909136   4.096  27.922   2.60481334   0.94243889  1.310851074545      0
jpeg:libjxl:d2.0      13270  1894603    1.1421485   3.831  22.792   3.18599176   1.12675147  1.286917548072      0
Aggregate:            13270  2919735    1.7601424   3.798  23.258   2.10445561   0.73980436  1.302161048110      0
```

Benchmark after:
```
Encoding            kPixels    Bytes          BPP  E MP/s  D MP/s     Max norm        pnorm       BPP*pnorm   Bugs
------------------------------------------------------------------------------------------------------------------
jpeg:libjxl:d0.5      13270  4604435    2.7757524   3.217  18.406   1.67270899   0.46916401  1.302283114386      0
jpeg:libjxl:d0.8      13270  3451612    2.0807808   3.843  21.973   1.71237743   0.62384150  1.298077432889      0
jpeg:libjxl:d0.9      13270  3189961    1.9230463   3.432  19.557   1.81622076   0.67879120  1.305346906713      0
jpeg:libjxl:d1.0      13270  2999278    1.8080944   3.483  20.767   1.95959663   0.71370775  1.290450959433      0
jpeg:libjxl:d1.1      13270  2801626    1.6889412   3.668  22.541   2.30924606   0.76993853  1.300380906241      0
jpeg:libjxl:d1.2      13270  2650997    1.5981355   3.619  22.952   2.31698585   0.81157827  1.297012067011      0
jpeg:libjxl:d1.5      13270  2280336    1.3746851   3.652  24.685   2.72100186   0.94873410  1.304210588039      0
jpeg:libjxl:d2.0      13270  1875238    1.1304745   3.880  27.784   3.36580706   1.13450375  1.282527537475      0
Aggregate:            13270  2887250    1.7405590   3.593  22.166   2.17430696   0.74545951  1.297516262872      0
```

* [butteraugli] Add support for writing the distmap to pfm.

* plugins/gimp: Support for decoding animations (libjxl#1863)

* plugins/gimp: Support for decoding animations

* lint

* fix authors check

* plugins/gimp: Handle the two blend modes GIMP understands.

* plugins/gimp: Handle named frames

* plugins/gimp: Fix maths fail

* plugins/gimp: Start layer index from 1 to match WebP/gif plugins

* set coalescing to true

Co-authored-by: Moritz Firsching <firsching@google.com>

* ssimulacra2: linear downscaling, fixed SSIM formula (libjxl#1848)

* Reapply the noise LUT fix from libjxl#1238 to the render pipeline (libjxl#1893)

* Use libjpeg's source manager api in streaming jpeg decoder.

This is a step towards having the same API as libjpeg.

* Use jpeg_decompress_struct fields for image dimensions.

* Make jpeg decoder's API more similar to libjpeg's API.

Modify ReadHeaders() to read until SOS marker as in libjpeg.

Add a StartDecompress() API function that will read whole
input for progressive images. This is the same behaviour as
in libjpeg's jpeg_start_decompress() for non-buffered-image mode.

* plugins/gimp: Fix progressive decoding (closes libjxl#1845 and libjxl#638) (libjxl#1870)

* Refactor WASM demo code

* Update dec_xyb.cc to fix wasm decode issue (libjxl#1900)

* Update dec_xyb.cc

Changed OR to AND. In WASM, images that were not XYB encoded would fail to decode, as `CanOutputToColorEncoding(c_desired)` is never called.

* fix authors check

Co-authored-by: Moritz Firsching <firsching@google.com>

* Fix the automatic setting of intensity_target from the input colorspace (libjxl#1899)

* Use Set(Quality|Distance) where appropriate

Goal: fix gcc-10 warning SetQuality is unused.

* Add demo site generator

* Add option to run custom cjpeg command in jpeg benchmark. (libjxl#1905)

With this addition, one can benchmark against a decoder-only libpeg
implementation by setting a custom LD_LIBRARY_PATH and using a
statically built cjpeg in the benchmark.

* Fix WASM test

* First version of decoder-only jpegli library. (libjxl#1909)

The jpegli library is a drop-in replacement to libjpeg62,
with improvements inspired by the JPEG XL codec.

The commit adds a minimal version of the libjpeg decoding API.

After building the jpeg and benchmark_xl targets, the following
command can be used to convert an input.jpg file to PNG:

```
LD_PRELOAD=./build/libjpeg.so.62 ./build/tools/benchmark_xl --input input.jpg --codec=jpeg --decode_only --save_decompressed --output_dir .
```

* Support 16 bit output buffer in libjpegli decoder. (libjxl#1913)

This can be enabled by the application by setting

```
  cinfo.quantize_colors = FALSE;
  cinfo.desired_number_of_colors = 65536;
```

In this case a scanline buffer should be twice as long.

Other changes in the commit:

  * Remove code duplication between input reading API functions.

  * Use the existing cinfo fields for marker parsing and keeping
    internal state, remove redundant fields from jpeg_decomp_master.

  * Add trivial implementation of jpeg_calc_output_dimensions
    and use cinfo->output_{width,height}, cinfo->out_color_components
    where appropiate.

* Update HWY

Add WASM-256 to build matrix

* Refactor ServiceWorker code

... to make it readable / hackable.

* libjpegli: reuse more fields of cinfo, fix build.

* mention Pale Moon support (libjxl#1915)

closes libjxl#1914

* Disable libjpegli on release and debian package builds.

* Avoid ODR violations

* Ensure alignment in OpHlg

* mention GraphicsMagick support (libjxl#1923)

* jpegli: fix decoding of non-interleaved sequential jpegs. (libjxl#1922)

* Add 16 bit jpeg output mode to benchmark_xl and cjxl. (libjxl#1918)

If the benchmark_xl and cjxl binaries are run with libjpegli,
then they will decode jpegs to 16 bit ppfs, otherwise they
work as before.

* Read cICP chunks from PNGs

- Where possible, create an encoded profile from the cICP values instead of
using an ICC profile.

- Change default rendering intent for decoded PNGs to "Relative".

- Honor the rendering intent specified by sRGB chunks when applicable.

* Explicitly set sRGB profile when reading the chunk

To ensure that sRGB takes priority when cHRM/gAMA occur earlier in the
file.

* fix hshift/vshift of metachannels (libjxl#1925)

* cjpeg_hdr: reuse EncodeJpeg

Co-authored-by: Luca Versari <veluca@google.com>

* Add a basic quality setting to cjpeg_hdr

Co-authored-by: Luca Versari <veluca@google.com>

* Fix NEON SIMD implementation of fast_lossless

Also enable SIMD in build scripts.

* remove deprecated or dead code

* jpegli: implement jpeg_consume_input()

Added tests for suspending source manager use-case, and tests for
consuming input ahead of the output processing.

* Refactor fast_lossless to support 16-bit data.

>8 bit paths are not SIMDfied yet.

* premultiply alpha in BufferToImageBundle (libjxl#1927)

* premultiply alpha in BufferToImageBundle

* remove premultiply_alpha arg

* remove arg also from SetAlpha

* don't include *_gbench.cc in code coverage report (libjxl#1939)

* Fix asan build.

* don't generate unused noise (libjxl#1943)

Fixes libjxl#1888

* Fix ExternalImageTest asan failure.

* Add an --update_on_failure flag to the conformance test runner.

Also update the conformance test set commit to make the tests pass.

* jpegli: implement buffered image mode

* jpegli: implement decoding to cropped output.

The implementation is not yet optimal, it just calls jpeg_read_scanlines
and discards the output that is not needed.

* Build release package for Ubuntu 22.04

* jpegli: implement raw data output mode

Drive-by: fix bug in bias computation for buffered image mode
and 32-bit architecture.

* Add libjpegli's libjpeg.so target to build/lib/jpegli directory.

* SIMDfy >8-bit fjxl on ARM NEON.

Benchmark on a A51, approximately a 10% speedup on a single core:

```
Before:

./fast_lossless artificial8.ppm x.jxl 2 100
    63.178 MP/s
     2.755 bits/pixel

./fast_lossless artificial13.ppm x.jxl 2 100
    32.670 MP/s
     6.994 bits/pixel

./fast_lossless artificial14.ppm x.jxl 2 100
    32.285 MP/s
     7.956 bits/pixel

./fast_lossless artificial16.ppm x.jxl 2 100
    27.592 MP/s
    10.411 bits/pixel

./fast_lossless bb8.ppm x.jxl 2 20
    37.628 MP/s
    11.895 bits/pixel

./fast_lossless bb13.ppm x.jxl 2 20
    21.407 MP/s
    26.594 bits/pixel

./fast_lossless bb14.ppm x.jxl 2 20
    20.894 MP/s
    29.654 bits/pixel

./fast_lossless bb16.ppm x.jxl 2 20
    18.621 MP/s
    35.989 bits/pixel

After:
./fast_lossless artificial8.ppm x.jxl 2 100
    63.353 MP/s
     2.755 bits/pixel

./fast_lossless artificial13.ppm x.jxl 2 100
    36.205 MP/s
     6.994 bits/pixel

./fast_lossless artificial14.ppm x.jxl 2 100
    34.875 MP/s
     7.956 bits/pixel

./fast_lossless artificial16.ppm x.jxl 2 100
    29.342 MP/s
    10.411 bits/pixel

./fast_lossless bb8.ppm x.jxl 2 20
    38.161 MP/s
    11.895 bits/pixel

./fast_lossless bb13.ppm x.jxl 2 20
    23.457 MP/s
    26.594 bits/pixel

./fast_lossless bb14.ppm x.jxl 2 20
    22.861 MP/s
    29.654 bits/pixel

./fast_lossless bb16.ppm x.jxl 2 20
    20.144 MP/s
    35.989 bits/pixel

```

Co-authored-by: Zoltan Szabadka <szabadka@google.com>
Co-authored-by: Leo Izen <leo.izen@gmail.com>
Co-authored-by: szabadka <9074039+szabadka@users.noreply.github.com>
Co-authored-by: alistair7 <alistair7@users.noreply.github.com>
Co-authored-by: Nigel Tao <nigeltao@golang.org>
Co-authored-by: Moritz Firsching <firsching@google.com>
Co-authored-by: Jon Sneyers <jon@cloudinary.com>
Co-authored-by: Sami Boukortt <sboukortt@google.com>
Co-authored-by: Luca Versari <veluca@google.com>
Co-authored-by: Fredrick Brennan <copypaste@kittens.ph>
Co-authored-by: Evgenii Kliuchnikov <eustas@google.com>
Co-authored-by: Jim Robinson <jimbo2150@gmail.com>
Co-authored-by: Damiano Albani <damiano.albani@gmail.com>

…#638) (libjxl#1870)