Make purge history slightly faster

Fix handling of redacted events from federation

* fix

* changelog

If we receive an event that doesn't pass their content hash check (e.g.
due to already being redacted) then we hit a bug which causes an
exception to be raised, which then promplty stops the event (and
request) from being processed.

This effects all sorts of federation APIs, including joining rooms with
a redacted state event.

Don't pull out events that are outliers and won't be deleted, as nothing
should happen to them.

expose number of real reserved users

Log outbound requests when we retry

Change the manhole SSH key to have more bits

fix VOIP crashes under Python 3

add some logging for the keyring queue

Fix jwt import check