Cap the time to retry txns to appservices to 8.5 minutes

Add some unit tests of prune_events()

Include typing events in initial v2 sync

Conflicts:
	synapse/handlers/sync.py

Add rooms that the user has left under archived in v2 sync.

Don't 500 on /login when the email doesn't map to a valid user ID.

There's been numerous issues with people playing around with their
application service and then not receiving events from their HS for
ages due to backoff timers reaching crazy heights (albeit capped at
< 1 day).

Reduce the max time between pokes to be 8.5 minutes (2^9 secs) which
is quick enough for people to wait it out (avg wait time being 4.25 min)
but long enough to actually give the AS breathing room if it needs it.

This is done by using the same /leave flow as you would use if you had
already accepted the invite and wanted to leave.

Repeatable unit test of event hashing/signing algorithm

Stuff signed data in a standalone object

Another signing test vector using an 'm.room.message' with content, so that the implementation will have to redact it

Add basic search API

Surely we don't need to preserve 'events_default' twice

Remove some login classes from synapse.

Makes both generating it in sydent, and verifying it here, simpler at
the cost of some repetition

Add config for how many bcrypt rounds to use for password hashes

Fix FilteringStore.get_user_filter to work with postgres

Implement third party identifier invites