[wgsl-in, wgsl-out, glsl-in] WebGPU compliant dual source blending feature

[Wumpf]

Connections

• Fixes Converge dual-source blending with the WebGPU standard #6402

Description

Makes the dual source implementation in wgpu WebGPU spec compliant.
This means changing the wgsl syntax & associated validation to match what's described here and here and here.
Furthermore, makes the dual source blending extension available when targeting WebGPU.

In order to reliably test wgsl output I incorrectly assumed that the way to go would be to add a glsl input to the snapshot tests (mostly to check for the enable directive being written out correctly), didn't realize that there's wgsl->wgsl snapshot testing as well, so I ended up implementing glsl index parsing as collateral😄

Testing
Added a slew of naga tests. Did manual testing in a pet project against native & WebGPU

Draft TODO

• [help wanted:] need to validate the the right enable directive is set but can't quite figure out how to access this in the right places
• [help wanted] [partially orthogonal]: While trying this out on my pet project I couldn't get it to fly in WebGPU because I was using naga_oil (with a hack, see also Support wgsl diagnostic directives. bevyengine/naga_oil#113 (comment)) where I have to rely on Module->write_wgsl which right now ignores all directives 😱 . Fixing that isn't entirely straight forward since Module is language agnostic so I have to detect the need for those
• a more vertical wgpu test checking that happy path works and that shaders needing wgsl require the device to have the correct feature set

Checklist

• Run cargo fmt.
• Run taplo format.
• Run cargo clippy. If applicable, add:
  • --target wasm32-unknown-unknown
• Run cargo xtask test to run tests.
• Add change to CHANGELOG.md. See simple instructions inside file.

[ErichDonGubler]

@Wumpf: Rebased and fixed up CI blockers. CI was unhappy because:

• clippy::unnecessary_map_or started firing for code authored before the CORE_MSRV upgrade in Upgrade CORE_MSRV to 1.82 and take advantage of some new features #7043.
• The GLSL test added here didn't enable the DUAL_SOURCE_BLENDING capability. I've enabled god_mode on the test to enable that capability, with a note of the intent of its usage.
  • We should iterate on breaking god_mode into more granular pieces. Having something that "enables all the things" sounds undesirable for tests exercising specific capabilities, shader stages, etc. CC @gfx-rs/wgpu on that.
  • We need to move away from opting into the capability at the API level, since reporting of the capability in standard WebGPU happens with ImplementedLanguageExtensions:all and is opted into per shader.

[cwfitzgerald]

We should iterate on breaking god_mode into more granular pieces. Having something that "enables all the things" sounds undesirable for tests exercising specific capabilities, shader stages

Had the same thought

[teoxoy]

Yeah, I think we should remove god_mode and just ask people to provide all required capabilities.

[cwfitzgerald]

Filed #7252