Support "worker queue" orginization of task processing

[rdpickard]

Currently tasks (fetch all mentions, loop over each mention, create response GIF in each iteration) processing combines the polling of Twitter and the running of the GoL simulation is in one loop. This is not conducive to "server-less" approach where each instance of the run is expected to be short lived, only for a min or two. See Heroku Scheduler or AWS Lambda documentation. An execution of one instance of the bot with the current implementation is likely to exceed the server-less period.

A worker queue organization where the polling of mentions is separated from the processing of GoL simulations will likely solve this.