core-code (#91) with necessary fixes

[joepvdbogaert]

Adds the following fixes to PR #91:

• Removed abstract decorators for implemented functions in the Predictor class.
• Changed the tuning method to return the minimum score instead of the maximum.
• Load the fullVisitorId as a string in the regression models notebook, so that they are read as intended.
• Removed the target variable and the ID from the data before fitting and predicting in the regression notebook.

In addition, I added a function that creates a submission file based on the chosen model and parameters and put it in a separate file: submit.py.

With these updates, I ran the entire pipeline, from the raw data to submission and got a reasonable score (without extensive tuning), so this should be a good starting point.

[joepvdbogaert]

Thanks @Christiannewisse, for the review. Can someone with write access approve now, then it can be merged (@steremma, @Dimkoim).