Occaisonally has_any_entity_changed would throw the error: "Set changed
size during iteration" when taking the max of the `sorteddict`. While
its uncertain how that happens, its quite inefficient to iterate over
the entire dict anyway so we change to using the more traditional
`bisect_*` functions.

Ensure we don't use unpersisted state group as prev group

Fix bug where state_group tables got corrupted

This is due to the fact that we prefilled caches using txn.call_after,
which always gets called including on error.

We fix this by making txn.call_after only fire when a transaction
completes successfully, which is what we want most of the time anyway.

Don't start user_directory handling on workers

Faster cache for get_joined_hosts

Split up device_lists_outbound_pokes table for faster updates.

Handle profile updates in user directory

Add a user directory

- Check if room is public when a user joins before adding to user dir
- Fix typo of field name "content.join_rules" -> "content.join_rule"