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

[Jeffrey Byrnes via llvm-commits]

jrbyrnes wrote:

> This pass, like most combiners, should work bottom up not top down. Would that avoid the issue?

We can use reverse iteration and deferred deletion to sort out the interactions between the two features. However, I think it is still best to leave the loop integration as a TODO at this point. Integrating them in the mentioned way causes seemingly nontrivial regressions in the lit tests caused by the integrated removeDeadInstrs.   

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