[clang] nonblocking/nonallocating attributes (was: nolock/noalloc) (PR #84983)

[Erich Keane via cfe-commits]

erichkeane wrote:

> Well assume attribute is a good example of where gcc and clang disagree and might cause issues. Gcc implemented the assume attribute just the same as the standard one. While clang implemented before standardization and now they differ and could cause issues. This is why I requested using clang_ here.

`assume` is an unfortunate situation, but it is also a particularly generic word (and we ended up being able to fix it fairly trouble-free).  In retrospect, we should have been more picky for a vendor-extension attribute instead of the OMP version.

So I'm not concerned about us being in the same situation here with `nolock/noalloc`.  In the future we/GCC should be better about communicating with each-other on these, but making some attribute spellings worse isn't worth the pain.

https://github.com/llvm/llvm-project/pull/84983