We should add a new error type that signifies to SPRAS that data is missing, to help with HTCondor error filtering, and potentially with #21.

Specifically, some kind of MissingDataError (as seen in #394) should be introduced to make SPRAS-internal errors and data requirement errors in generate_inputs distinct from one another.