[llvm] [RISCV] Allow non-loop invariant steps in RISCVGatherScatterLowering (PR #122244)

[Philip Reames via llvm-commits]

https://github.com/preames commented:

I *think* the reasoning is okay here because having a strided start + splat(A) + splat(B) + splat(C) is still a strided pattern.  I'm a bit concerned about the logic for unwinding through non-recurrence adjustments, but I *think* this is correct.  I'd like to see a bit more targeted testing (in particular, remove EVL!) to convince myself this is generally correct.

https://github.com/llvm/llvm-project/pull/122244