Add more feature preprocessing operators

[rhiever]

There are some feature preprocessors that could potentially work well in the TPOT environment:

• Binarizer
• OneHotEncoder
• MaxAbsScaler
• MinMaxScaler

Allowing TPOT to work with these preprocessors shouldn't be much more expensive, and it will allow TPOT to explore different ways of processing the features.

[rhiever]

We won't be able to add support for OneHotEncoder until #29 is solved. I will add support for the other three preprocessors.

Further complication with OneHotEncoder: It should only be applied to categorical columns. Thus, we need an effective heuristic to detect categorical vs. continuous columns. I've raised a question on StackOverflow to discuss ideas: http://stackoverflow.com/questions/35826912/what-is-a-good-heuristic-to-detect-if-a-column-in-a-pandas-dataframe-is-categori

[rhiever]

The three preprocessing operators are now implemented in this branch: https://github.com/rhiever/tpot/tree/more-feature-preprocessors

TODO:

• Add export() support for them
• Add them to the docs