[llvm] SCEV: teach isImpliedViaOperations about samesign (PR #124270)

[Ramkumar Ramachandra via llvm-commits]

artagnon wrote:

> Isn't the symbolic-BTC more refined? Have to think about why this introduces a regression.

After some more investigation, it appears that SCEVExpander's `isHighCostExpansion` returns false now, and the LFTR in IndVarSimplify kicks in and replaces things. Still have to figure out how to avoid the regression.

https://github.com/llvm/llvm-project/pull/124270