[llvm] Reapply [AMDGPU] Avoid resource propagation for recursion through multiple functions (PR #112251)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm commented:

It seems the precedent is for MCExprs to [not support recursion](https://github.com/llvm/llvm-project/blob/5c1752e368585e55c0335a7d7651fe43d42af282/llvm/lib/MC/MCParser/AsmParser.cpp#L6403)

If we are going this way, this code should be unified to deal with the weak edge cases.

Alternatively it feels more natural to directly support recursion in expressions, but maybe that is more difficult. 

https://github.com/llvm/llvm-project/pull/112251