Separate the bytes written by masked moves for tracing/taint tools

Somehow I thought we were already doing this: but it looks like we do not expand or specially handle masked moves such as x86's `maskmovq` for tracing or taint tracking tools. Since it's a contiguous address range it's not as significant to ignore the mask as for scatter-gather: but it could conditionally affect a couple of cache lines or none for simulation, and byte-level taint tracking will propagate the wrong data.

For simulation, having the value of the mask inside the simulator would be an alternative to unrolling the instruction into bytes, but unrolling is the more general solution as it will work for taint tracking as well.