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

[Pranav Gorantla via llvm-commits]

pranav-159 wrote:

Thanks for the feedback. I'll rework this to match on the ISD::VECREDUCE_* nodes directly and handle the lowering in LowerVECREDUCE instead. 

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