[llvm] [ConstraintElim] Check of other OP is guaranteed to not be undef/poison. (PR #76182)

[Florian Hahn via llvm-commits]

fhahn wrote:

Yeah, I checked and it looks like in all cases selects are already transformed to bitwise ops in practice, so this isn't needed.

I am wondering if it would be worth in general to verify some canonical form assumptions at certain points, to catch cases where canonicalization where missed.

https://github.com/llvm/llvm-project/pull/76182