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

[Doug Wyatt via cfe-commits]

dougsonos wrote:

> aha! I missed this was part of the plan to begin with, that makes sense.

Well, there was some head-scratching before debugging revealed that the Clang was treating `malloc()` as a built-in. It doesn't do that if you declare it locally.

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