[clang] [llvm] [LangRef] Clarify specification for float min/max operations (PR #172012)

[via cfe-commits]

ActuallyaDeviloper wrote:

>     * In the X86 backend, **delete the code for minnum lowering** and always treat minnum like minimumnum, which already implements all the required signed zero and nsz handling. The implementation complexity for this change for the X86 backend is literally negative.
* not quite, it's a must to utilize `VRANGE` for `minnum` afterwards from my point of view. It's a 1-1 mapping with and without `nsz` so it should be rather trivial.

https://github.com/llvm/llvm-project/pull/172012