[llvm] [InstSimplify] Optimize maximumnum and minimumnum (PR #139581)

[Lewis Crawford via llvm-commits]

LewisCrawford wrote:

Here is the NVPTX maximumnum ISel patch I mentioned above: [#155804](https://github.com/llvm/llvm-project/pull/155804)

I think we'll need to land this InstSimplify patch, the above ISel patch, and maybe a couple of other improvements like better SelectionDAG expansion of v2f16 `maximumnum` calls in ISel before we'd be fully ready to use `llvm.maximumnum` for `nvvm.fmax`, but we're getting closer.

https://github.com/llvm/llvm-project/pull/139581