[clang] [clang][analyzer] Improved PointerSubChecker (PR #93676)

[Balazs Benics via cfe-commits]

https://github.com/steakhal commented:

I'd say it looks pretty good. Objectively speaking it improves the TP rate.
I have compared these reports against the constant interpreters of clang and gcc to see if in constexpr context which of these expression would trigger "UB" there.
I found I think only 3 places where this implementation disagrees what I think we should report.

https://github.com/llvm/llvm-project/pull/93676