Prohibit copyTextureToBuffer from compressed textures.
Prohibit copyTextureToTexture where either source or destination is a compressed format.
(copyBufferToTexture copies are ok.)

This is restriction #3 from the compat proposal.

`GPUTextureViewDescriptor.dimension` must not be `"cube-array"` in
Compatibility Mode (issue #4 from the compatibility-mode proposal).

Merge from 'main' at 20b5f62

At bind group creation time, validate that texture views contain all
array layers (view.baseArrayLayer == 0, view.arrayLayerCount ==
texture.depthOrArrayLayers).

This is [restriction
#6](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#6-array-texture-views-used-in-bind-groups-must-consist-of-the-entire-array-that-is-basearraylayer-must-be-zero-and-arraylayercount-must-be-equal-to-the-size-of-the-texture-array)
from the compatibility-mode proposal.

Prevent the use of rg32float, rg32sint, and rg32unit as STORAGE_BINDING
in Compatibility Mode.

Remove the line about createShaderModule() validation, since it's
unnecessary.

This is [restriction
#8](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#8-disallow-two-component-rg-texture-formats-in-storage-texture-bindings)
in the Compatibility Mode proposal.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>

During encoder bind group validation, validate that all
`GPUTextureViews` in all bind groups containing a reference to the same
`GPUTexture` must have matching `baseMipLevel`, `mipLevelCount`, and
`aspect`.

This validation represents [restriction
#5](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#5-views-of-the-same-texture-used-in-a-single-draw-may-not-differ-in-aspect-or-mip-levels)
in the Compatibility Mode proposal.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>
Co-authored-by: Kai Ninomiya <kainino@chromium.org>

GPUDepthStencilState.depthBiasClamp must be zero in Compat mode.

This represents [restriction
9](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#9-depth-bias-clamp-must-be-zero)
from the Compatibility Mode proposal.

Source and destination `sampleCount` must be 1.

This is [issue
#12](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#12-disallow-copying-multisample-textures)
in the Compatibility Mode proposal.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>

Add the Compatibility Mode limits to the limits table.

Increase the `colspan` of the full-width rows to match the new column
count (4 -> 5).
Implement the Compat validation of maxVertexAttributes, in which `vertex_index` and `instance_index` each consume a slot.
This represents [restriction
10](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#10-lower-limits)
from the compatibility-mode proposal.

---------

Co-authored-by: Mehmet Oguz Derin <mehmetoguzderin@mehmetoguzderin.com>

All members of viewFormats must be equal to the texture's format.

This is achieved by changing the definition of "texture view format
compatible" to be strict equality in Compatibility mode.

This corresponds to [restriction
13](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#13-disallow-texture-format-reinterpretation)
in the compatibility-mode proposal.

---------

Co-authored-by: Kai Ninomiya <kainino@chromium.org>

Sync compat feature branch for CI

This represents restriction
[23](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#23-disallow-all-multisampled-integer-textures)
in the compatibility-mode proposal.

- bgra8unorm-srgb
- multisampled r32float and rgba16float.

This represents issues
[15](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#15-disallow-bgra8unorm-srgb-textures)
and
[22](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#22-disallow-multisampled-rgba16float-and-r32float-textures)
in the proposal doc.

Also modify the proposal doc to validate this at pipeline creation, not
shader module creation.

This represents issue
[7](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#7-disallow-sample_mask-and-sample_index-builtins-in-wgsl)
in the proposal doc.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>

Add `textureBindingViewDimension` to `GPUTexture` and `GPUTextureDescriptor`.

Implement the defaulting algorithm for
`GPUTextureDescriptor.textureBindingViewDimension`.
Implement Compat-specific validation in `createBindGroup()` validation.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>
Co-authored-by: Kai Ninomiya <kainino@chromium.org>

In Compatibility Mode, disallow interpolation type "linear", and
interpolation sampling "sample" and "first".
When interpolation type is "flat", allow only interpolation sampling
"either".

This represents restrictions

[11](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#11-disallow-linear-and-sample-interpolation-options)
and

[17](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#17-disallow-interpolationflat-and-interpolationflat-first)
in the compatibility-mode proposal.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>

Implement the depth texture restictions for Compatibility Mode:

- no `textureLoad()` of depth textures
- only comparison samplers with depth textures in texturing functions

This represents restrictions

[16](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#16-disallow-textureload-with-texture_depth-textures)
and

[20](https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#20-disallow-using-a-depth-texture-with-a-non-comparison-sampler)
from the compatibility-mode proposal.

Follow-up to #5297

2D textures must have a single layer.
Cube textures must have 6 layers.

This represents issue

[14](https://github.com/SenorBlanco/gpuweb/blob/featurebranch-compat/proposals/compatibility-mode.md#14-require-depthorarraylayers-to-be-compatible-with-texturebindingviewdimension-in-createtexture)
in the compatibility-mode proposal.

Implement new storage buffer and texture limits.
    
Implement the new limits:
- maxStorageBuffersInVertexStage
- maxStorageBuffersInFragmentStage
- maxStorageTexturesInVertexStage
- maxStorageTexturesInFragmentStage
    
This represents restrictions

[18]https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#18-introduce-new-maxstoragebuffersinvertexstage-and-maxstoragetexturesinvertexstage-limits
and

[19]https://github.com/gpuweb/gpuweb/blob/main/proposals/compatibility-mode.md#19-introduce-new-maxstoragebuffersinfragmentstage-and-maxstoragetexturesinfragmentstage-limits
in the compatibility-mode proposal.

Limit maximum texture/sampler combinations
    
This represents restriction

[11](https://github.com/SenorBlanco/gpuweb/blob/featurebranch-compat/proposals/compatibility-mode.md#21-limit-the-number-of-texturesampler-combinations-in-a-stage)
in the compatibility-mode proposal.

Following up on #5361, this PR adds
`swizzle` to the texture views match algorithm for compatibility mode.

Define an internal slot on the adapter for "default feature level".
Use it to store the requested feature level (if supported).
Use it to determine whether an adapter is "Core-defaulting" or "Compatibility-defaulting".
Set the "core-featres-and-limits" feature in the device as appropriate.

---------

Co-authored-by: Kai Ninomiya <kainino1@gmail.com>

This started with addressing François's editorial comment to define
"Compatibility Mode", but I determined that most of the references to
"Compatibility Mode" should actually be more descriptive and point more
precisely to relevant definitions, and that the definitions of the
feature level strings should be more comprehensive.

Also removed definitions of "Core-defaulting" and
"Compatibility-defaulting" in favor of direct links (but kept those
phrases in the prose as they're useful for understanding).

Additionally I noticed that were missing text to make
`"core-features-and-limits"` increase the limits, and text to explicitly
say we set the default limits depending on feature level. Fixed those.

Issue:
#5402 (review)