[llvm] [NVPTX] Generalize and extend upsizing when lowering 8/16-bit-element vector loads/stores (PR #119622)

[Drew Kersnar via llvm-commits]

dakersnar wrote:

Done and done. I think Alex is taking a quick look, but otherwise this should be good to go.

https://github.com/llvm/llvm-project/pull/119622