[llvm] [polly] [InstSimplify] Fix incorrect poison propagation when folding phi (PR #96631)

[Nikita Popov via llvm-commits]

nikic wrote:

Unfortunately, we just found the first instance where this miscompile occurs in the wild (https://github.com/llvm/llvm-project/issues/97702).

Per the comments of @RalfJung in https://discourse.llvm.org/t/rfc-add-nopoison-attribute-metadata/79833 it looks like Rust does not want to use `nopoison` attributes, which basically leaves us no choice but to move forward with this as-is and have Rust eat the resulting codegen regressions.

https://github.com/llvm/llvm-project/pull/96631