This extends SCEV verification to check not only backedge-taken counts, but all entries in the IR -> SCEV cache. The restrictions are the same as for the BECount case, i.e. we ignore expressions based on undef, we only diagnose constant deltas (there are way too many false positives otherwise) and we limit to reachable code.
Not entirely sure whether doing this is worthwhile, as it's probably quite expensive, and I didn't catch any new issues in llvm-lit tests / llvm-test-suite at least.