[llvm] [AMDGPU] Move tryEffectiveStall to GCNSchedStrategy class, optionally use in tryPendingCandidate (PR #191924)

[Kevin Choi via llvm-commits]

choikwa wrote:

> Shouldn't the test show what the option/`tryEffectiveStall` is doing? Also, why is this behind an option at all?

I'm told there were some regressions when tryPendingCandidate was allowed to consider latency in the past. I want to limit the consideration of latency in the pending queue to scenarios such as DS_READs taking longer than usual. I'm hoping subsequent PRs will be able to show an improvement in the scheduling output of the test.

https://github.com/llvm/llvm-project/pull/191924