[llvm] [LV] Add stricter checks on potential induction PHIs (PR #75866)

[Aleksandr Popov via llvm-commits]

aleks-tmb wrote:

> I haven't taken a close look at this, but it could be LV's usage of SCEV needs fixing rather than limiting the vectorizer?

Maybe it's better to limit the vectorization that's **incorrect** at this point and support the case later in the SCEV, since it might take more time?

https://github.com/llvm/llvm-project/pull/75866