Previously if what could be a load, store, branch, or call instruction
was located within last two bytes of the program image being
disassembled, the disassembler would try to read out of bounds to get
its address parameter. This change marks such cases as not being valid
instructions.
Previously address arguments were extracted without verifying the start
byte corresponded to a valid instruction. In case of a non-instruction
byte this would result in the two following bytes to be silently
skipped.