[llvm] [NVPTX] fold movs into loads and stores (PR #144581)

[Alex MacLean via llvm-commits]

https://github.com/AlexMaclean commented:

One thing I wonder about is how this changes the PTX semantics of a program. If we change from loading something as a b32 to a v2.b16 will this impact the memory consistency guarantees in PTX? 

https://github.com/llvm/llvm-project/pull/144581