Hierarchy change: take Black-box optimization out of RL

[schaul]

Although it technically fits there, it is a bit confusing.
I think the split should be along the difference of ontogenetic/phylogenetic with on one side optimization, evolution, pso, etc. (coevolution methods should fit here, but how about multi-objective optimization?) and on the other side policy gradients, and other RL algos.

[schaul]

Thomas says:

There was a class "Learner", which we thought of as being the RL-equivalent of "Trainer" for supervised learning. Then it was changed to "RLLearner" and another class "Learner" was created that does optimization now. For me it always felt like the blackbox optimization code was just dumped in there because some functionality was similar to the existing RL methods. We should clean this up or work out a proper common framework for RL and optimization.

[schaul]

Additionally, we could change the directory hierarchy a bit more: merge the tools and auxiliary folders, put datasets in there too, and move the test folder one level up.

[schaul]

I've started doing those changes in a new branch: http://github.com/schaul/pybrain/commits/learnerstucture ...