AVX-512: Fix 1-bit mask opcodes and possibly introduce 1-bit mask size type.

Some AVX-512 instructions read only 1 mask bit (e.g. vmovss etc.) To better describe them, we may need a sub-1-byte type to describe a bit. Currently, we're using KEb to describe them.

The opcodes vaddss, vsubss, vmulss are wrong and need to be the new type, or KEb. Thanks @JivanH for finding.