Catch any exceptions in the pusher loop. Use a lower timeout for pushers so we can see if they're actually still running.

Add script to (re)convert the pushers table to changing the unique key. Also give the python db upgrade scripts the database engine so they can convert parameter strings, and add *args **kwargs to the upgrade function so we can add more args in future and previous scripts will ignore them.

Thumbnail images on a seperate thread

Improve backfill.

Don't needlessly compute prev_state

Don't compute EventContext for outliers.

Add cache to get_state_groups.

SYN-403: Make content repository use its own http client.

Add config option to disable compression of http responses

Before, caches were global and so different instances of the stores
would share caches. This caused problems in the unit tests.

contrib/systemd: log_config.yaml: do not disable existing loggers

It turned out that merely configuring the root logger is not enough for
"catch-all" semantics. The logging subsystem also needs to be told not
to disable existing loggers (so that their messages will get propagated
to handlers up the logging hierarchy, not just silently discarded).

Signed-off-by: Ivan Shapovalov <intelfx100@gmail.com>

Allow endpoint for verifying recaptcha to be configured