Fix stray comma in function signatures in Metal

[teoxoy]

Connections
Fixes Bug 1992500

Description
Fixes a regression introduced by #8265 where we were mistakenly writing a comma before the first buffer argument (if no other arguments were printed before it).

Testing
Enabled more CTS tests.

Squash or Rebase?
Rebase.

[cwfitzgerald]

Does this need backporting?

[ErichDonGubler]

@cwfitzgerald: I'm not @teoxoy, but I'll say: yesplz

[andyleiserson]

As far as we know it's only a CTS regression, right? I think backporting can at least wait for input from @teoxoy.

[teoxoy]

I think newer Metal versions are fine with the stray comma since I couldn't repro this issue on my M1 either.

[teoxoy]

I can put up another PR with the backport though.

[teoxoy]

@cwfitzgerald I pushed on the remote instead of the origin (my fork) so it ended up directly in the v27 branch... is that ok? Or should I reset, force push and still open a separate PR?

[cwfitzgerald]

That's fine. I can bump version and release

[cwfitzgerald]

@teoxoy seems ci failed on v27. Could you investigate?

[ErichDonGubler]

@cwfitzgerald: I don't think this PR caused the CI failure; all failures report:

[INFO  wgpu_xtask::cts] Reading default test list from cts_runner/test.lst
[INFO  wgpu_xtask::cts] Fetching CTS
[INFO  wgpu_xtask::cts] Checking out CTS
fatal: unable to read tree (5e7bd6ed86201123ff6b0900974587550afb10e7)
Error: Failed to check out CTS

[andyleiserson]

2ea52e0

(The problem is that I bumped the CTS version on trunk earlier today, and the CTS is cloned with --depth 1. So the cache from trunk will not work for the v27 branch. Certainly there's a better fix but changing the tag should suffice for now.)

[ErichDonGubler]

@andyleiserson: Filed #8313.