When we do an invite rejection, save the signed leave event to the db

During a rejection of an invite received over federation, we ask a remote
server to make us a `leave` event, then sign it, then send that with
`send_leave`.

We were saving the *unsigned* version of the event (which has a different event
id to the signed version) to our db (and sending it to the clients), whereas
other servers in the room will have seen the *signed* version. We're not aware
of any actual problems that caused, except that it makes the database confusing
to look at and generally leaves the room in a weird state.

Accept join events from all servers

Make sure that we accept join events from any server, rather than just the
origin server, to make the federation join dance work correctly.

(Fixes #1893).

Cache glob to regex at a higher level for push

The cache wrappers had a habit of leaking the logcontext into the reactor while
the lookup function was running, and then not restoring it correctly when the
lookup function had completed. It's all the fault of
`preserve_context_over_{fn,deferred}` which are basically a bit broken.

Fix bug where current_state_events renamed to current_state_ids

* Bail early if remote wouldn't be retried

* Don't always return true

* Just use get_retry_limiter

* Spelling

Don't user upsert to persist new one time keys

Notify on new federation traffic

Fix installation issues

Batch sending of device list pokes

Short circuit if all new events have same state group

Use presence replication stream to invalidate cache

Add slave transaction store to workers who send federation requests

Instead of using the cache invalidation replication stream to invalidate
the _get_presence_cache, we can instead rely on the presence replication
stream. This reduces the amount of replication traffic considerably.

Cache hosts in room

try not to drop context after federation requests

Add another missing yield on check_device_registered

Reduce some CPU work on DB threads

cursor_to_dict can be surprisinglh expensive for large result sets, so lets
only call it when we need to.