1: drop CpuVREX
2: don't abort() upon DATA16 prefix on (E)VEX encoded insn
3: correct permitted operand sizes for AVX512 scatter/gather
4: add more exhaustive tests for invalid AVX512 zeroing-masking
5: fold various AVX512 templates with so far differing Masking attributes
6: also optimize KXOR{D,Q} and KANDN{D,Q}
Jan