Add estimators parameter

Adding an "estimators" Papermill parameter for enabling controlling the total run time when testing.

AMLS Pipeline Release

This extends the previous implementation by adding a notebook that creates an Azure Machine Learning Service Pipeline that does the work of notebooks 04, hyperparameter tuning, and 05, creating the best model.

Papermill patch

Upgrade to Papermill 1.0.1 to fix a version mismatch with nbconvert

Train-tune-test

This major update supports a train-valid-test split of the data per Issue #4. It includes:

A training script that uses train and valid data, and logs running statistics of model performance on valid.
A test script that uses a downloaded model to score the test data.
A new notebook that uses the best hyperparameters to train a model using a large number of estimators.