[compiler-rt] [compiler-rt] Also consider SIGPROF as a synchronous signal (PR #85188)

[Nazım Can Altınova via llvm-commits]

canova wrote:

@dvyukov
> This looks wrong. SIGPROF is generally not synchronous. Handling it synchronously will cause memory corruptions. The right way to deal with such deadlocks is to fix why the thread is not delivering it later. Usually a thread is blocked in some syscall that is not marked as BLOCK_REAL in tsan runtime.

Marking functions as blocking with `BLOCK_REAL` is the exact thing I wanted to do initially (in #83844 and #83561). And I actually sent a PR (#84162) to fix one of the issues I encountered, but it got no attention/feedback so far. I also asked a question on #83844 to learn on how to intercept syscalls (as oppose to intercepting glibc calls, we need to intercept FUTEX syscalls for this).

https://github.com/llvm/llvm-project/pull/85188