Setup release-plz

[Firestar99]

• https://release-plz.dev/
• see test run creating a release PR for this branch
  • PRs created by release-plz DO NOT trigger ci due to github limitations, do we want that?
• configured so changelogs must we written manually, since we don't adhere to conventional commits
  • I think for rust-gpu this is necessary anyway, since most changes we do are not user-facing, so an auto-generated changelog would likely confuse users more than help
  • if PRs were merged or squashed, we could extract the PR titles from the commits, but since we just rebase that's not an option
• trusted publishing is set up
• Added a dry-run of a release to CI
  • this job shows a build failure due to me excluding important header files from packaging
  • takes ~15min to run, even when failing, that's ~4.5 times slower than a build. Specifically a failing build taking 15min+ may indicate something in release-plz taking forever before packaging and testing the release?

Notes for porting this to rust-gpu:

• github's merge strategy and merge queue seem to need special attention
• alpha releases are supported, just set the workspace version to an alpha and it'll increment the alpha version
• I've observed release-plz only changing the changelog of spirv-tools-sys and leave spirv-tools alone, seemingly wanting to release only that one since it's the only one with changes. However, it bumped the version in the workspace, also bumping the version of spirv-tools without touching the changelog. During development, something made it suddenly start updating the changelog of spirv-tools as well, I can't exactly pinpoint what. This could cause issues with rustc_codegen_spirv and spirv-std always needing to be the same version.

[Firestar99]

just discovered there's a dry-run, want to add that since this repo is particularly sensitive to packaging due to selectively including cpp and header files

[Firestar99]

I don't know why in the last commit the first release-plz-dry-run failed, but the second one succeeded. I'll call it a fluke for now and we'll see if it shows up again.

@LegNeato ready for review, I'd like to publish this crate to see if everything works before copying this to rust-gpu.

[LegNeato]

lgtm