Filter out very low-significance SubGRBTargeted events

Currently we have an incredible flux of search='SubGRBTargeted' events, especially from Fermi, that have FARs larger than the max allowed values. We have previously just applied the NOT_GRB label to prevent these from sending alerts, but much of our analysis requires manual followup of EM_COINC events, which isn't occurring because of the high numbers of these SubGRBTargeted events oversaturating this channel. Fixes include, from least to most severe:

• Run analysis but don't apply EM_COINC label for events that have an insufficient FAR, and therefore not part of an official search
• Submit these as group='Test' events so they are store for future analysis but no further analysis is done
• Block these events from being uploaded altogether