[llvm] AMDGPU: VIMAGE/VSAMPLE are bundle candidates (PR #140854)

[via llvm-commits]

ruiling wrote:

> No objection to the patch, but I don't understand why this pass is useful. The original description (from https://reviews.llvm.org/D72737) said that we bundle the clauses formed by the pre-RA scheduler to prevent the post-RA scheduler from breaking them, _instead of_ adding clustering mutations to the post-RA scheduler. But something has changed since then, and we _do_ add clustering mutations to the post-RA scheduler.

One of the reasons would be that pre-ra and post-ra scheduler uses different versions of `tryCandidate`. We would do aggressive clustering like in max-memory-clause schedule strategy. So this pass is useful like in the case.

https://github.com/llvm/llvm-project/pull/140854