[X86] combinePredicateReduction - split vXi16 allof(cmpeq()) to vXi8 allof(cmpeq())

vXi16 patterns allof(cmp()) reduction patterns will have to be pack the comparison results to vXi8 to use PMOVMSKB.

If we're reducing cmpeq(), then we can compare the vXi8 halves directly - similar to what we already do for vXi64 -> vXi32 for cases without PCMPEQQ.