Test default layout bindgroups

[greggman]

It's probably a good idea to decide what we want to happen here and fix it ASAP because as it is, Chrome will allow bind groups made with different layout: 'auto' pipelines to work with each other but the spec seems to say that's disallowed. It's easy to imagine lots of WebGPU pages are already breaking this rule.

I also suspect the spec would like to say that all empty bind group layouts are compatible but I don't see that in the spec currently so it's tested that they are not compatible in this PR.

---

Requirements for PR author:

• All missing test coverage is tracked with "TODO" or .unimplemented().
• New helpers are /** documented */ and new helper files are found in helper_index.txt.
• Test behaves as expected in a WebGPU implementation. (If not passing, explain above.)

Requirements for reviewer sign-off:

• Tests are properly located in the test tree.
• Test descriptions allow a reader to "read only the test plans and evaluate coverage completeness", and accurately reflect the test code.
• Tests provide complete coverage (including validation control cases). Missing coverage MUST be covered by TODOs.
• Helpers and types promote readability and maintainability.

When landing this PR, be sure to make any necessary issue status updates.

[kainino0x]

The spec rules as written only require it to come from the same pipeline, not the same slot:
https://gpuweb.github.io/gpuweb/#bind-group-compatibility
This is a little odd, but probably fine as I don't think it's much of a footgun?

[greggman]

In that case we need to test that 2 bind groups that appear the same, are the same? In other words

      @group(0) @binding(0) var s1: sampler;
      @group(0) @binding(1) var t1: texture_2d<f32>;

      @group(1) @binding(0) var s2: sampler;
      @group(1) @binding(1) var t2: texture_2d<f32>;

If we use layout: 'auto' on this then a bindGroup made with pipeline.getBindGroupLayout(0) should work bound to bind group 1.

I don't think there is a test for that and it would be easy for browsers to disagree on this without a test?

It does make me wonder why they aren't just compatible in general. The rules are spelled out for what gets generated.

[kainino0x]

good point, we would definitely need to test that. Unless we want to make a breaking change in the spec and make it stricter.

The exclusivePipeline thing was added as a way to prevent footguns because auto layouts are sensitive to static uses in WGSL which are fairly subtle and easy to change (like by commenting out a function call, it removes from the BGL some binding that was used by that function's call tree). But I do wonder if it's really needed. Maybe with some good error messages there would be no footgun, like, "bindgroup created with auto BGL from pipeline 'X' but is used with pipeline 'Y' which has an incompatible auto BGL. Note that auto BGLs do not include declared bindings which are not reachable from the entrypoint of the pipeline".
Jasper and Brandon both had more opinions about this. We definitely picked the current behavior knowing it was a more "safe for 1.0" decision that we didn't necessarily have to keep forever.