[llvm] [LV] Make greater use of ScalarEvolution in addDiffRuntimeChecks (PR #213668)

[David Sherwood via llvm-commits]

david-arm wrote:

> Could you make sure we have a test with a nested loop, where the entry to the scalar loop is != the outer header? In this case the invariant computations should be hoisted out of the loop nest by the SCEV expander

Done!

https://github.com/llvm/llvm-project/pull/213668