Add the CACHEDIR.TAG file to the cache directory

[perillo]

Implement the http://www.brynosaurus.com/cachedir/ specification for the Zig cache directory.

This change will make it easier to exclude Zig cached data from a backup, expecially when the directory is not a well-known cache directory, like ~/.cache/zig or zig-cache.

Closes #8210

[perillo]

For @r00ster91: I don't why I don't see your comment here (maybe you deleted it?). I used the link http://www.brynosaurus.com/cachedir/ instead of the actual link after a redirect (https://bford.info/cachedir/) because this is what the specification suggested.

While it is true that the specification has not been updated since 2004, some implementations like Kopia use the original link.

[perillo]

On Linux the CI scripts fail with unimplemented. I'm not sure if this is related to this changeset.

[perillo]

The error from the CI is

[15/21] Running zig1.wasm to produce /home/ci/actions-runner1/_work/zig/zig/build-new/compiler_rt.c
FAILED: compiler_rt.c 
cd /home/ci/actions-runner1/_work/zig/zig && /home/ci/actions-runner1/_work/zig/zig/build-new/zig1 /home/ci/actions-runner1/_work/zig/zig/lib build-obj lib/compiler_rt.zig -ofmt=c -OReleaseSmall --name compiler_rt -femit-bin="/home/ci/actions-runner1/_work/zig/zig/build-new/compiler_rt.c" --mod build_options::/home/ci/actions-runner1/_work/zig/zig/build-new/config.zig --deps build_options -target x86_64-linux
unimplemented
Aborted
[16/21] Running zig1.wasm to produce /home/ci/actions-runner1/_work/zig/zig/build-new/zig2.c
FAILED: zig2.c 
cd /home/ci/actions-runner1/_work/zig/zig && /home/ci/actions-runner1/_work/zig/zig/build-new/zig1 /home/ci/actions-runner1/_work/zig/zig/lib build-exe src/main.zig -ofmt=c -lc -OReleaseSmall --name zig2 -femit-bin="/home/ci/actions-runner1/_work/zig/zig/build-new/zig2.c" --mod build_options::/home/ci/actions-runner1/_work/zig/zig/build-new/config.zig --deps build_options -target x86_64-linux
unimplemented

Should the tagCacheDirectory function be a no op when the target is wasm?

Thanks.

[andrewrk]

@Vexu I don't think this should be merged as-is because it should not block the rest of the compilation pipeline. I think it should be done as an asynchronous job that gets completed by performAllTheWork().

Anyway, closing this now since it's an abandoned PR that never passed the CI checks. @perillo can ask for help in the issue corresponding to this PR. I like to keep the PR queue filled with only patches that are considered ready for review & merge by their authors.