[llvm] [AMDGPU] Remove SIWholeQuadMode pass early exit (PR #98450)

[Carl Ritson via llvm-commits]

perlfu wrote:

> Seems reasonable but it would be easier to review if you can separate out the refactoring, the bug fix and the optimization.

I'll separate the patches.
I did some initial compile time testing on an early version of this without the exact mode lowering bypass and the impact was 0.03% increase in compile time -- I think this version is probably lower.

https://github.com/llvm/llvm-project/pull/98450