[llvm] [LV] Add vplan folds for urem(X, PowerOf2) -> and(X, PowerOf2 - 1) (PR #212198)

[David Sherwood via llvm-commits]

david-arm wrote:

> This has caused [stage2-O3 +0.11%](https://llvm-compile-time-tracker.com/compare.php?from=293d7c559bdf21c9739d3c389d9a4932da461620&to=4b0b8cbac0f567cdacfde0c81019fc6c0d27f251&stat=instructions:u) compile-time regression. I don't think that's intentional?

Definitely not intentional. I wasn't expecting such a minor change to cause such a large compile-time regression, especially given how many folds already exist in vplan.

https://github.com/llvm/llvm-project/pull/212198