0.8
- auto-unregistring of dead/slow workers
- introduced a graceful shutdown
- added a worker registering so restarting workers are not impacting the system
- drastically reduced the number of used FDs, mainly by reusing the same context when possible
- the broker exits if there's already a valid broker running in the socket.
- powerhose-broker gained 2 new options: --check and --purge-ghosts
- implemented timeout_max_overflow in the client.
- the stacks are dumped on worker timeouts
- now using delayed callbacks for the heartbeat
- make sure execution errors are properly transmited and raised.
- fixed the pool of connectors - so every connector is correctly freed
- now workers can get extra options from the command-line
- initial release.