Fix user erasure and re-enable

This reverts commit 1d009013.

This reverts commit ce0d9111, reversing
changes made to b4a5d767.

Fix event_push_actions tablescan when reinserting events

Add some logging to /search and /context queries

Make _get_joined_hosts_cache cache non-iterable

Make push actions rotation configurable

Support hiding events from deleted users

Simplify get_persisted_pdu

remove dead filter_events_for_clients

Improve StateGroupCache efficiency for wildcard lookups

When _get_state_for_groups is given a wildcard filter, just do a complete
lookup. Hopefully this will give us the best of both worlds by not filling up
the ram if we only need one or two keys, but also making the cache still work
for the federation reader usecase.

Optimise state_group_cache update

Log number of events fetched from DB

so that we can stub it for the sentinel and not have a billion failing UTs

Fix description of "python_gc_time" metric

When we finish processing a request, log the number of events we fetched from
the database to handle it.

[I'm trying to figure out which requests are responsible for large amounts of
event cache churn. It may turn out to be more helpful to add counts to the
prometheus per-request/block metrics, but that is an extension to this code
anyway.]

Write a clear restart indicator in logs

Add error code to room creation error

I'm fed up with never being able to find the point a server restarted in the
logs.

SECURITY UPDATE: Prevent unauthorised users from setting state events in a room
when there is no `m.room.power_levels` event in force in the room. (PR #3397)

Discussion around the Matrix Spec change proposal for this change can be
followed at https://github.com/matrix-org/matrix-doc/issues/1304.

Adjust event auth rules when there is no PL event

We need power levels for this test to do what it is supposed to do.