[llvm] AMDGPU: Remove flat/global fmin/fmax intrinsics (PR #105642)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> Can you provide a similar intrinsic which can be called as Intrinsic::int_amdgcn_global_atomic_fmax which calls your desired intrinsic instead?

No. The point is no intrinsics. Adding an identical intrinsic helps nothing. Only atomicrmw is an acceptable atomic representation 

https://github.com/llvm/llvm-project/pull/105642