fix validation of dual source blending structs

[vxpm]

Description
validation for structs that contain a blend_src attribute in one of their members is wrong. the WebGPU spec says:

For each structure type S defined in a WGSL module (not just those used in shader stage inputs or outputs), let members be the set of members of S that have location attributes.

If any entry in members specifies a blend_src attribute, members must use dual source blending, which means:

• members must contain exactly 2 entries, one with @location(0) @blend_src(0) and one with @location(0) @blend_src(1).
• All the members must have same data type.

Otherwise, members must not contain two entries with the same location value.

(source)

the part where naga is wrong is this one:

let members be the set of members of S that have location attributes

currently, naga uses all the members of S during validation, instead of just the members with location attributes. this means you can't use any members with builtin attributes (such as frag_depth).

this PR fixes this by only checking members with location attributes.

related PR: #8856

[andyleiserson]

I think it would make sense to implement the validation of blend_src structs in valid/type.rs. The spec says to do the validation:

for each structure type S defined in a WGSL module (not just those used in shader stage inputs or outputs)

Then, I think we could remove some of the interface validation related to blend_src and rely on the type validation.

There are some CTS tests related to dual-source blending that we may be able to pass after this change (there's information on running CTS tests in docs/testing.md):

• webgpu:api,validation,render_pipeline,fragment_state:dual_source_blending,*
• webgpu:shader,validation,extension,dual_source_blending:*

The only other check I'm aware of that's missing is something like this, with the color target validation in create_render_pipeline:

if dual_source_blending && color_targets.len() > 1 {
    return Err(
        pipeline::CreateRenderPipelineError::DualSourceBlendingWithMultipleColorTargets {
            count: color_targets.len() as u32,
        },
    );
}

[andyleiserson]

The error message text uses the labels blend_src(0) and blend_src(1), but location_members could have them in the opposite order if they were declared that way in the source.

I think it would be possible to do one pass over the members and save the types of the blend_src members in an array [Option<Handle<Type>>; 2] along the way using the actual index values, so that they are identified correctly in the error message.