One room at a time so we don't take out the whole server with leave
events, and restart at server restart.

This implements this very crudely: this probably isn't viable
because parting a user from all their rooms could take a long time,
and if the HS gets restarted in that time the process will be
aborted.

notifications: Convert next_token to string according to the spec

Fix metrics that have integer value labels

Currently the parameter is serialized as an integer.

Signed-off-by: Konstantinos Sideris <sideris.konstantin@gmail.com>

Make 'unexpected logging context' into warnings

Fix logcontext leaks in rate limiter

Fix 'Unhandled Error' logs with Twisted 18.4

fix http request timeout code

Refactor event storage to prepare for changes in state calculations

Make Client-Server API return 403 for invalid token

Escape label values in prometheus metrics

This gets two arguments, not one.

Fix a class of logcontext leaks

Fix logcontext leak in media repo

Fix a couple of logcontext leaks in unit tests

Make FileResponder.write_to_consumer uphold the logcontext contract

... which were making other, innocent, tests, fail.

Plus remove a spurious unittest.DEBUG which was making the output noisy.

So, it turns out that if you have a first `Deferred` `D1`, you can add a
callback which returns another `Deferred` `D2`, and `D2` must then complete
before any further callbacks on `D1` will execute (and later callbacks on `D1`
get the *result* of `D2` rather than `D2` itself).

So, `D1` might have `called=True` (as in, it has started running its
callbacks), but any new callbacks added to `D1` won't get run until `D2`
completes - so if you `yield D1` in an `inlineCallbacks` function, your `yield`
will 'block'.

In conclusion: some of our assumptions in `logcontext` were invalid. We need to
make sure that we don't optimise out the logcontext juggling when this
situation happens. Fortunately, it is easy to detect by checking `D1.paused`.

Fix incorrect reference to StringIO

This was introduced in 4f2f5171