[llvm] [AMDGPU] Add block carried latency to CoExecSched (PR #192324)

[Austin Kerbow via llvm-commits]

================

----------------
kerbowa wrote:

We don't bound this scan downstream. We do bound some other lookahead scans, but I suggest leaving this one unbounded unless we find real cases where it becomes a compile-time bottleneck like the other scans turned out to be.

https://github.com/llvm/llvm-project/pull/192324