improve agc_match insertion

On crowded fields (frameid=755735, say, with ~850 objects), the agc_match insert takes ~10s. The ag.py loop calls data_utils.py:write_agc_match(), which builds a numpy array which is then passed down to agActor/opdb.py:insert_agc_match() which iterates over the rows and inserts them one-by-one with the regular obdb.insert(). Replace those with some bulk insert. Umm, build a pandas DataFrame and pass that to the same opdb.insert()? That is OK, and is used elsewhere.

For most fields this is not a big issue.