Add notebook using TensorFlow to perform the predictions

[volpatto]

In this PR, a notebook using TensorFlow to perform the predictions through Neural Networks is provided (synced with a related .py script).

The notebook has 4 main parts:

• Data preprocessing: loading, splitting, feature engineering, etc.
• Hyperparameter (Adam's learning rate) tuning: a 4-folds Randomized Search is employed through sklearn.
• Train/Test stage: using the optimal learning rate, train/test two NNs: one with GHS constraints and other unconstrained.
• Results: several plots assessing the outcomes are provided. Results are checked for each std. property individually. A thermodynamical consistency check is done at the end of the notebook.

Click here to see the rendered notebook.

[volpatto]

• Fix all deps versions in the devenv file