[clang] [analyzer] Remove some false negatives in StackAddrEscapeChecker (PR #125638)

[Michael Flanders via cfe-commits]

Flandini wrote:

> Thanks a lot for the change and persisting. Sorry for the stupid questions :) Looks good to me.

No worries at all, I am happier with the PR state now compared to before. Thanks for the review.

I've added more test coverage so I believe we have similar test coverage as Sema does for the `-Wreturn-stack-address`. 

I found one false negative we have compared to Sema on `musttail` attribute return statements and added some test cases for these. I think this would be better in a different PR since it is not about the return value, more about lifetimes than stack address escape checking, and more about the SVals in the return expression syntax tree.

@Xazax-hun, could you also merge this PR for me?

https://github.com/llvm/llvm-project/pull/125638