Make the preprocessing script operational

[NathanRxl]

As a beginning, the preprocessing script must be able to run with the command

& python3-conda preprocessing_script.py

And must create from csv files "data/train.csv" and "data/test.csv" new csv files "data/preprocessed_train.csv" and "data/preprocessed_test.csv".

The preprocessing must be using the strategy we discussed last Thursday, that is to say :

• NaN are replaced with -1 for non categorical features.
• Categories are encoded with target mean for categorical features, NaN being considered as one of the possible categories.

The assignee should tackle any easy(/free)-improvement of the given description on-the-fly.