[clang] nonblocking/nonallocating attributes: 2nd pass caller/callee analysis (PR #99656)

[Doug Wyatt via cfe-commits]

dougsonos wrote:

I wish my locally-built clang format, from my branch (up-to-date with main a few days ago) behaved the same as the one in automation :(  Not sure what to do.

https://github.com/llvm/llvm-project/pull/99656