This should be a little more accessible an entry point for client side stuff then staging Galaxy jobs as an example.

Includes a generalizations for how dynamic outputs are collected from the working directory, this was going to be needed to implement discovered datasets on the Galaxy side anyway.