[llvm] Revert "[AMDGPU] Examine instructions in pending queues during schedu… (PR #166201)

[Ravil Dorozhinskii via llvm-commits]

ravil-mobile wrote:

@antiagainst, we need this for Triton. We need to adjust Triton based on the MLIR upstream changes and run the CI 

https://github.com/llvm/llvm-project/pull/166201