[llvm] [ConstraintElimination] Handle equality predicates with signed system (PR #205331)

[Florian Hahn via llvm-commits]

https://github.com/fhahn commented:

Overall seems fine, although it would be interesting why there are some quite notable regressions with LTO, which is why originally the queries are kept separate. I'm personally fine with the increase in compile-time, just wondering if there's something we could do better to soften the blow with LTO.

Although I don't think there's much we can do, as it is probably just due to larger systems when doing LTO

https://github.com/llvm/llvm-project/pull/205331