[llvm] [InstCombine] Allow folds of shifts by constants for scalable vectors again (PR #132522)

[Luke Lau via llvm-commits]

lukel97 wrote:

> Can we generally accept scalable splats in m_ImmConstant? I doubt this is the only place with this problem.

I've reworked m_ImmConstant to accept scalable splats, so the changes in InstCombineShifts are gone now. 

I had to create two separate structs to help match it, because one binds and the other doesn't (and I couldn't seem to reuse cstval_pred_ty because it's not polymorphic and needs to be specicialized to either ConstantInt or ConstantFP).

It was tempting to try and create one binding matcher and declare a dummy Constant for the non-binding m_ImmConstant(), but it looks like everything here is trying to be tail callable so I avoided it.

https://github.com/llvm/llvm-project/pull/132522