The influx writer currently rejects the whole block of listens if influx rejects a single bad value in the set of listens, which will cause correct listens to not be recorded. As we did for the redis_pubsub (sigh) we should do this for the influx_writer and examine if this needs doing for the bigquery-writer.

Also, we should add an integration test to ensure that this doesn't happen. (import a block of 2 listens, 1 good 1 bad; in the end 1 should end up in the DB)