Add partial derivative builtins

[evahop]

Adds the following from gfx-rs/wgpu#4402

• dpdxCoarse
• dpdxFine
• dpdyCoarse
• dpdyFine
• fwidthCoarse
• fwidthFine

Feel free to make any changes btw.

[evahop]

Clippy is freaking out for me locally, so I'm just force pushing now that I've appeased the CI

running cargo clean resolved most of my issues, however I still get:

error: lint `clippy::derive_hash_xor_eq` has been renamed to `clippy::derived_hash_with_manual_eq`
   --> src/back/msl/sampler.rs:159:9
    |
159 | #[allow(clippy::derive_hash_xor_eq)]
    |         ^^^^^^^^^^^^^^^^^^^^^^^^^^ help: use the new name: `clippy::derived_hash_with_manual_eq`
    |
    = note: `-D renamed-and-removed-lints` implied by `-D warnings`

error: could not compile `naga` (lib) due to previous error

Which even the CI suggested, however it then complained after I made the switch, and then didn't complain when I reverted it. 😵‍💫

[teoxoy]

Thanks for the PR!
Overall it looks good!

One nit: Please try to to avoid using _ for match arms especially when that coincides with one remaining enum variant.

[teoxoy]

misclicked

[evahop]

Thanks for the PR! Overall it looks good!

One nit: Please try to to avoid using _ for match arms especially when that coincides with one remaining enum variant.

Sounds good, does this also apply to resting the pattern if there's only one remaining field { x , .. } vs { x, y: _ }?

[teoxoy]

Sounds good, does this also apply to resting the pattern if there's only one remaining field { x , .. } vs { x, y: _ }?

I think that's fine and depends on context; sometimes it might be worth explicitly ignoring members, sometimes we might not care at all and use the rest operator.

[teoxoy]

A rebase should take care of the linting issues; since we landed #2279.

[teoxoy]

Looks great - thank you for the improvements!