[llvm] [X86] Narrow vXi32/vXi64 usubsat to vpsubusb/vpsubusw when LHS is known to fit in fewer bits (PR #206592)

[Simon Pilgrim via llvm-commits]

================

----------------
RKSimon wrote:

To make this more generic, wouldn't this be better as a search loop, starting from 8 (or ScalarBits  / 2) and doubling/laving until you find a legal match with the right amount of active bits?

https://github.com/llvm/llvm-project/pull/206592