Iterating through a number of possible converters is not very efficient. Since we generate value holders and they're corresponding converters, the encoders for pipeline operators, stages, etc. can leverage a map that maps the value holder class name to its converter, allowing for more efficient encoding of generated pipelines.