Fix sql error in quarantine_media

Signed-off-by: Travis Ralston <travpc@gmail.com>

Remove spurious log argument

Remove unused "event_type" param on state.get_current_state_ids

Improve exception handling in persist_event

Logging and metrics for the http pusher

Use a connection pool for the SimpleHttpClient

Signed-off-by: Travis Ralston <travpc@gmail.com>

Signed-off-by: Travis Ralston <travpc@gmail.com>

... which would cause scary-looking and unhelpful errors in the log on dns fail

this param doesn't seem to be used, and is a bit pointless anyway because it
can easily be replicated by the caller. It is also horrible, because it changes
the return type of the method.

1. use `deferred.errback()` instead of `deferred.errback(e)`, which means that
a Failure object will be constructed using the current exception state,
*including* its stack trace - so the stack trace is saved in the Failure,
leading to better exception reports.

2. Set `consumeErrors=True` on the ObservableDeferred, because we know that
there will always be at least one observer - which avoids a spurious "CRITICAL:
unhandled exception in Deferred" error in the logs

Fix SQL for user search

fix some syntax errors for user search when search_all_users is enabled

fixes #2801, hopefully

Make it possible to run tests against postgres

... instead of creating our own special SQLiteMemoryDbPool, whose purpose was a
bit of a mystery.

For some reason this makes one of the tests run slightly slower, so bump the
sleep(). Sorry.

Configure the connectionpool used for unit tests to run the `on_new_connection`
function.

Remove unused/bitrotted MemoryDataStore

Factor out get_db_conn to HomeServer base class

Fix server 500 on public rooms call when no rooms exist

This function is identical to all subclasses, so we may as well push it up to
the base class to reduce duplication (and make use of it in the tests)

This isn't used, and looks thoroughly bitrotted.

rather than try reconstruct the results object, better to guard against the xrange step argument being 0

synapse 500s on a call to publicRooms in the case where the number of public rooms is zero, the specific cause is due to xrange trying to use a step value of zero, but if the total room number really is zero then it makes sense to just bail and save the extra processing

Add some comments about the reactor tick time metric