some ugly fixes in the file_service for improving panda output from spreadsheet processing that I need to revist.
now that the spiff-workflow handles multi-instance, we can't have random multi-instance tasks around.
Improved tests around study deletion.