[llvm] [NVPTX] support packed f32 instructions for sm_100+ (PR #126337)

[Nikolay Panchenko via llvm-commits]

npanchen wrote:

there's one more runtime issue I later hit due to that change. [Here are](https://gist.github.com/npanchen/f207c79206795caaba470d6395aabbfb) 2 ptxes (bad, good) with and without that change. I'm by no means expert in PTX and to me packs, unpacks are correct. @Artem-B @Prince781 @AlexMaclean can you please take a look?

https://github.com/llvm/llvm-project/pull/126337