Becca

This version of BECCA contains some major changes to both the deep learning algorithm (the ziptie) and the reinforcement learner (the cerebellum). It works better than ever on the benchmark.

There are a couple of new changes in this release. The drivetrain has been simplified. Gearboxes have been eliminated. Zipties are simpler. Most importantly, zipties have been sped up by around 10X by implementing the most time intensive parts in numba. As a result of this and other improvements, the drivetrain is making beautiful feature hierarchies in both [becca_world_watch](https://github.com/brohrer/becca_world_watch) and [becca_world_mnist](https://github.com/brohrer/becca_world_mnist).

I didn't bother to update the documentation in the wiki this time around, so some of it will be outdated. I wanted to be sure to tag this commit so that there would be a reliable restore point for anyone trying to implement the watch or mnist worlds.

This version of BECCA sports a reworked reinforcement learning algorithm. There are significant changes to both the hub and arborkey. In its new form, only actions can be selected as goals. It used to be that features could be selected as goals as well, but, despite its theoretical appeal, this didn't help BECCA learn faster or perform better on its current task set. Curiosity has been implemented in the hub in a way that tries to walk the exploration/exploitation line a little more elegantly. The arborkey now includes restlessness, a propensity to act when the agent hasn't acted for a while. This helps the agent keep from getting stuck in long term ruts of poor performance.

The benchmark has grown to include another task, a version of the 1D grid task in which the reward is delayed by a small random number of time steps. This tests an agent's ability to handle a challenging credit assignment problem appropriately.

[See BECCA in action.](http://youtu.be/uU1_13c6umo?list=PLF861CC4C40439EEB)

This is a major release. BECCA's core reinforcement learning algorithms have been completely reworked. The hub is largely unchanged, but several other pieces have been added to it: **spindle** (attention), **mainspring** (short and long term declarative memory), and **arborkey** (a top level executive). Blocks are now named **gearboxes** and together they form a **drivetrain**.

This is a minor release. The learning algorithms are unchanged since 0.5.0, but some minor changes were introduced in order to make BECCA compatible with the Robot Operating System, ROS.

This is a major release. BECCA's core algorithms have been completely reworked. Notably, the signal clustering algorithm **ziptie** and the sequence creation algorithm **daisychain** are new, as is goal selection algorithm in the **hub**.