[llvm] AMDGPU: Add round-to-odd rounding during f64 to bf16 conversion (PR #133995)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm commented:

Can you clean up the description to refer to just the changes here, without referring to the out of tree commit context 

https://github.com/llvm/llvm-project/pull/133995