Fix for splitting with outputs_to_working_directory = True, also rename a dict and method in jobs to refer to hda instead of dataset to avoid (my own) confusion.