[llvm] [TTI] getCommonMaskedMemoryOpCost - consistently use getScalarizationOverhead instead of ExtractElement costs for address/mask extraction. (PR #87771)

[David Sherwood via llvm-commits]

https://github.com/david-arm approved this pull request.

LGTM! I think this looks correct and more accurate logically than before, since the insert/extract costs could be different depending upon the lane. I think if some costs have gone up, then it's up to the target to fix getVectorInstrCost to be more accurate.

https://github.com/llvm/llvm-project/pull/87771