Automatically restart Unicorn workers based on 1) max number of requests and 2) max memory