[llvm] [AMDGPU] Insert readfirstlane for uniform VGPR arguments (PR #178198)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> after this, are you able to address this quickly or should we temporarily revert? (Or do we not revert for updates to experimental arches?)

We don't revert or report updates for experimental targets 

https://github.com/llvm/llvm-project/pull/178198