DataFlow-level "Don't save output" options

[eschenfeldt]

"Don't save output" can be a nice compromise to increase efficiency while maintaining the traceability of still having the runs themselves stored but it can be tedious to apply it to a lot of modules.

A couple possible options:

• "Only Save External" - external inputs and outputs are saved but all other outputs are suppressed
• "None" - all outputs are suppressed
• "Smart" - apply some heuristics to past runs to identify large outputs (or maybe even try to add a heuristic for whether an output is redundant if we wanted to get ludicrously fancy).