[llvm] [X86] Interleave independent reductions to share cross-lane extracts (PR #209963)

[Simon Pilgrim via llvm-commits]

RKSimon wrote:

Thanks for looking at this - we're in the process of moving x86 to add support for ISD::VECREDUCE_* nodes (and avoid all usage of matchBinOpReduction) - can you look at performing this expansion as part of LowerVECREDUCE instead?

https://github.com/llvm/llvm-project/pull/209963