[llvm] [AMDGPU][True16][CodeGen] workaround sgpr folding in true16 (PR #128929)

[Joe Nash via llvm-commits]

https://github.com/Sisyph commented:

In the commit title I would change workaround to 'Implement'
LGTM besides that and comment change.
I will let someone else approve it. 


https://github.com/llvm/llvm-project/pull/128929