Use musl's pthread_detach

[sbc100]

The musl version handles that case where pthread_detach is called
on a thread that is in the middle of exit'ing.

[juj]

This comment is about pthread_exit? Should it instead say about pthread_join?

[juj]

we get
a code size reduction

How big of a difference was this for code size?

This PR should definitely be run against https://github.com/juj/posixtestsuite to verify that the new implementation is still conformant, or if not, check if there is a good reason not to be. Pthread detach tests are at https://github.com/juj/posixtestsuite/tree/master/conformance/interfaces/pthread_detach

[sbc100]

OK, I'll hold off until I can at least run the testsuite.

Surely we need to integrate that testsuite somehow? Otherwise we can't know if any of our changes are safe right? Should I open a bug to do that?

[sbc100]

Also, assuming we can find a test that fails, but also fails in upstream musl, can we agree that that is acceptable. i.e. there is no need for us to be more conformant the musl itself? I think that same statement is mostly true for the rest of libc too.

[sbc100]

Adding posixtestsuite in #12923

[sbc100]

Reviving this patch. We now run posixtest suite are part of CI and I also ran it locally to confirm everything passes. I did have to patch in a couple of extra checks that musl was overlooking.

[juj]

lgtm!