Refactor all memory protocol decodes to have a common base class, as well as output data types

Anything that ends up producing a sequence of "read value from this address" / "write value to this address" should be able to integrate with common visualizations, e.g. "dump state of memory at this point in time" or "temporal hexdump" visualization a la https://scanlime.org/2009/10/temporal-hex-dump/