[llvm] [AMDGPU] Add IR LiveReg type-based optimization (PR #66838)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm approved this pull request.

Don't know why this is in AMDGPULateCodeGenPrepare instead of regular AMDGPUCodeGenPrepare. IIRC it was an extremely niche issue that caused us to have this late one in the first place which I wanted to fix. We can always move it later though 

https://github.com/llvm/llvm-project/pull/66838