[llvm] [X86] Improve FREEZE node elimination for SETCC operations (PR #192362)

[Simon Pilgrim via llvm-commits]

RKSimon wrote:

If we're confident the source node types to setcc/carry can't create poison, and we drop the canCreateUndefOrPoisonForTargetNode override I think we should be OK - isGuaranteedNotToBeUndefOrPoisonForTargetNode can't get us stuck in the same way (it'll just immediately fold freeze(setcc()) -> setcc()). 

https://github.com/llvm/llvm-project/pull/192362