[llvm] [X86]: Rewrite demorgan rule for ANDN (PR #163789)

[via llvm-commits]

kper wrote:

Let's the disregard the fact that I had to update some test files needlessly for now, e.g. `llvm/test/CodeGen/PowerPC/xxeval-and-nand.ll` and `llvm/test/CodeGen/PowerPC/vec_veqv_vnand_vorc.ll`

Looking at `llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll`, 
`llvm/test/CodeGen/X86/bool-ext-inc.ll`, `llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll`, `llvm/test/CodeGen/PowerPC/xxeval-eqv-nor-or-xor.ll`, the three operand variant does not seem very profitable as well. I had also to remove the snippet `llvm/lib/Target/X86/X86ISelLowering.cpp` to avoid a loop.

https://github.com/llvm/llvm-project/pull/163789