incremental_orbitfit fails unless the id column is named provID

[matthewholman]

_observations_for_update does not take primary_id_column_name and reads
d["provID"] directly (orbitfit.py:1735 and :1747), so the sequential-update
path only works when the id column happens to have the default name:

incremental_orbitfit(data, cache_dir, prior, primary_id_column_name="ObjID")
# ValueError: no field of name provID

via sequential_update -> _observations_for_update. orbitfit() handles
ObjID fine, so only the incremental path is affected. The existing incremental
tests all use provID, which is why it does not show up.

Fix is to thread the column name through and use it.