Add some meaningfull training to example jupyter notebook

[YevheniiSemendiak]

Currently, our example notebook only performs dataset load from public to storage.
We should replace it / expand to some meaningful training, which could be tracked by the MLFlow attached to Jupyter.