[llvm] [SelectionDAG] Avoid illegal wide vectors expanding get.active.lane.mask (PR #213670)

[Kito Cheng via llvm-commits]

kito-cheng wrote:

> How come we're seeing get.active.lane.mask being generated on RISC-V? I thought the loop vectorizer always emitted a step-vector + icmp on RISC-V unless it was overridden by the `-force-tail-folding-style` flag

The case was found by @preames, and it can't reproduced/vectorized by upstream any more, but I think the point is it's valid IR but crash.

opt (20): https://godbolt.org/z/4ovMP67ET 
llc: https://godbolt.org/z/snrx8EM7G

https://github.com/llvm/llvm-project/pull/213670