[llvm] [SCEV] Handle backedge-count logic for std::reverse like loops (PR #92560)

[Eli Friedman via llvm-commits]

efriedma-quic wrote:

> For future reference, what other cases do you think we could handle here?

I think you can special-case the case where the positive and negative strides are equal: if the arithmetic doesn't overflow immediately, it can't overflow at all.  Maybe you can generalize that a bit...

https://github.com/llvm/llvm-project/pull/92560