[llvm] [InstCombine] Fold umul.overflow(x, c1) | (x*c1 > c2) to x > c2/c1 (PR #147327)

[Marius Kamp via llvm-commits]

mskamp wrote:

I've pushed twice for an easier review: The first push comprises the changes, the second push does the rebase.

https://github.com/llvm/llvm-project/pull/147327