<!-- The PR description should answer 2 important questions: -->

### What

In hasura/v3-engine#1954 we started using
`authorization_rules` internally for type output permissions. Here, we
do the same for select permissions for models, using
`authorization_rules` in `plan` to calculate:
- can we select from this `Model`?
- are there any filters we need to apply?
- are we allowed to do subscriptions with it?
- which argument presets do we need?
- which relational mutations are we allowed (`update`, `delete`,
`insert`)?

### Performance

Benchmarks is measured against current `main`.

`Total Execution Time` is a bit slower for simpler queries, but faster
for more complex queries, particularly those using relationships.

This tracks with the changes as:
- we cache the results of evaluating `Conditions` to optimise for
running `x-hasura-role == foo` checks over and over again, but we are
still doing more work than a simple map lookup.
- when collecting filters for `ModelPermissions`, we return
`Vec<&ModelPredicate>` and pass that around, only cloning them when
actually needed. This reduces a bunch of unnecessary cloning,
particularly where relationships are concerned, as we'd check a user
could access the model underneath a relationship but never use the
filters for anything else.

```
simple_select/bench_execute/Total Execution time
                        time:   [648.73 µs 659.03 µs 669.90 µs]
                        change: [+11.312% +14.294% +17.206%] (p = 0.00 < 0.05)
                        Performance has regressed.

select_many/bench_execute/Total Execution time
                        time:   [608.95 µs 619.06 µs 629.69 µs]
                        change: [+8.5431% +10.842% +13.304%] (p = 0.00 < 0.05)
                        Performance has regressed.

select_many_where/bench_execute/Total Execution time
                        time:   [1.2079 ms 1.2287 ms 1.2506 ms]
                        change: [+0.0203% +2.1676% +4.6082%] (p = 0.05 < 0.05)
                        Change within noise threshold.

object_relationship/bench_execute/Total Execution time
                        time:   [831.13 µs 843.36 µs 855.92 µs]
                        change: [-0.4508% +1.6093% +3.5083%] (p = 0.12 > 0.05)
                        No change in performance detected.

array_relationship/bench_execute/Total Execution time
                        time:   [779.65 µs 792.49 µs 806.21 µs]
                        change: [-17.943% -15.474% -12.910%] (p = 0.00 < 0.05)
                        Performance has improved.

relay_node_field/bench_execute/Total Execution time
                        time:   [605.03 µs 614.19 µs 623.73 µs]
                        change: [-29.244% -26.453% -23.661%] (p = 0.00 < 0.05)
                        Performance has improved.
```

---------

Co-authored-by: Brandon Simmons <brandon.m.simmons@gmail.com>
V3_GIT_ORIGIN_REV_ID: 6d96e5ccbdd6c095112c58cda430ce30b01b3544