[llvm] [RISCV] Lower vector_reverse for zvfhmin/zvfbfmin (PR #110218)

[Luke Lau via llvm-commits]

lukel97 wrote:

> The vrgaher change here seems related to any shuffle we lower to a vrgather, is there something which makes reverse special here?  I'd rather see this tested via normal shuffle lowering if not.  

Just a quick update on this, I'm working on splitting out the vrgather_vl patterns but this involves lowering build_vector for bf16 to test it, which involves a few other small fixes. 

Most ops aren't lowered yet on fixed length vectors, so far I've only be focusing on scalable vectors. 

https://github.com/llvm/llvm-project/pull/110218