[llvm] Reapply "[LV] Use ExtractLane(LastActiveLane, V) live outs when tail-folding. (#149042)" (PR #168738)

[Luke Lau via llvm-commits]

lukel97 wrote:

I've collected some performance results for llvm-test-suite and SPEC CPU 2017 on `-march=x86-64-v3 -O3 -flto` and there didn't seem to be any regressions: https://lnt.lukelau.me/db_default/v4/nts/1238?compare_to=1237
One of the interleaving microbenchmarks benchForIC2VW1LoopWithReductionTC15 improved, presumably because it can be tail folded now.

https://github.com/llvm/llvm-project/pull/168738