Separate federation servlet into different lists

Fix caching error in the push evaluator

Initialising `result` to `{}` in the parameters meant that every call to
_flatten_dict used the *same* target dictionary.

I'm hopeful this will fix https://github.com/matrix-org/synapse/issues/2270,
but I suspect it won't. (This code seems to have been here since forever,
unlike the bug, and I don't really think it explains the observed
behaviour). Still, it makes it hard to investigate the problem.

Increase default cache size

Improve the error handling for bad invites received over federation

Prefill forward extrems and event to state groups

Cache macaroon parse and validation

Turns out this can be quite expensive for requests, and is easily
cachable. We don't cache the lookup to the DB so invalidation still
works.

Use an ExpiringCache for storing registration sessions

This is because pruning them was a significant performance drain on
matrix.org

Make workers report to master for user ip updates

Batch upsert user ips

Reject local events that don't round trip the DB

Fix up indices for users_who_share_rooms

Add configuration parameter to allow redaction of content from push m…

Add configuration parameter to allow redaction of content from push messages for google/apple devices

Store URL cache preview downloads separately

This makes it easier to clear old media out at a later date

Fix thinko in initial public room user spam