Implement softmax activation function for the output layers

• Replace the Input / hidden / output setter and getter function, with a generic function that allow to change the size of the layer. https://github.com/AndreaCatania/brain/blob/master/brain/brain_areas/uniform_brain_area.h#L35
• Try to see if is possible to improve the current organization of Weights, Biases, Activation, without adding too much information. (The idea is to make the code more clear and easy to manage)

Creates something that track the execution of the NEAT.

This is important to understand what's going on during the execution of the different epochs.

(An idea is to pass an object to the advance_epoch function that fill it, in this way anyone can treat the data as he wish)

Please implement the derivative of softmax in the Uniform brain area

In this moment the SharpBrainArea is using a function recursion to get the output.
IMO this is really good idea, but since the stack memory is not infinite the structure size is limited, and this limit change depending on the hardware where is running.

For this reason, even if the execution performances (at least with not too much recursion), I would like to re-implement it using a Loop.

https://github.com/AndreaCatania/brain/blob/neat/brain/brain_areas/sharp_brain_area.cpp#L163-L191