[wgsl] Integrate the fragment_depth proposal into the spec - #6299

Previews, as seen when this build job started (ebb7fe5):
WebGPU _{webgpu.idl | Explainer | Correspondence Reference}
WGSL _{grammar.js | wgsl.lalr.txt}

I think the next steps are committee review of the PR and CTS tests before this can land.

Re: Mike's request for an idea of benefits.

I got a useful response from gemini by asking "what is conservative depth in graphics rendering".

Excerpting from the reply:

##Common Use Cases
You will typically see conservative depth used in advanced rendering techniques where geometry shape is faked or adjusted inside the pixel shader:

• Raymarching & Interior Mapping: Simulating complex 3D volumes (like clouds, liquids, or deep windows on a flat wall) inside a simple bounding box.

• Screen-Space Decals: Mapping bullet holes or posters onto uneven geometry where the depth needs slight adjustment to prevent clipping or z-fighting.

I* mpostors / Billboards: Rendering complex 3D assets (like distant trees or particles) on simple 2D cards while pushing the depth back to correctly match the shape of the real 3D asset.

• Conservative Rasterization Matching: Used alongside conservative rasterization to make sure pixel depths align properly when calculating things like exact voxel grids or highly accurate shadow maps.

In terms of prevalence/reach:

It became an industry-wide standard alongside the release of DirectX 11 (2009) and OpenGL 4.2 (2011). If you are running hardware released in the last decade or more, it fully supports conservative depth.

It should benefit at least these GPUs:

• NV since GTX 400
• AMD GCN and RDNA GPUs
• Intel Arc and Xe

When we last discussed this in committee, I was the only person with remaining concerns, and they're addressed now. It's got a full list of approvals, so I think we can merge this without bringing it up in committee again.

I need to write CTS tests I believe before this can land in the spec. I think that's the only missing piece at this point.