Tag events with a background ID

For reproducability, we should tag events with some kind of background_id, relating to which version of marginalized stats was used to calculate FAR.

It could be a hash, a count of how many times calcfap has run, the start/end time of the marginalized stats, etc.

The main point is to be able to identify why a sudden change in FAR occurs, both within a run & between different runs.

Current workarounds are to run injection/noninjection tests on fixed backgrounds, and possibly in #11 we'll make calcfap block.