[llvm] [NVPTX] Don't use stack memory when bitcasting to/from v2i8 (PR #113928)

[Justin Fargnoli via llvm-commits]

================
@@ -0,0 +1,36 @@
+; RUN: llc < %s -mtriple=nvptx64-nvidia-cuda -mcpu=sm_90 -mattr=+ptx80 -asm-verbose=false \
----------------
justinfargnoli wrote:

Since we're already checking for most of the PTX that are generated for each function, I'd recommend auto-generating the `CHECK` statements. 

https://github.com/llvm/llvm-project/pull/113928