Rearrange persist_event so that do all the queries that need to be done before returning early if we have already persisted that event.

Keep around the old (buggy) version of the prune_event function so that we can use it to check signatures for events on old servers

Try to ensure we don't persist an event we have already persisted. In persist_event check if we already have the event, if so then update instead of replacing so that we don't cause a bump of the stream_ordering.

Don't fail an entire request if one of the returned events fails a signature check. If an event does fail a signature check, look in the local database and request it from the originator.

Rejections

Client v2 sync

Add doc string for __nonzero__ overrides for sync results, raise not implemented if the client attempts to do a gapless sync

Conflicts:
	synapse/storage/schema/im.sql