Mule has problems reading timeseries in fieldsfiles. Would be better to have a separate function to remove these? Something similar to `um_fields_subset.py` or something mule_based (if the read error can be trapped and handled).