support refit in cross_validation

[jmoralez]

Adds the refit argument to NeuralForecast.cross_validation which can be a boolean or an integer and works in the following way:

• If False, the models are only trained at the start.
• If True, the models are trained on every window.
• If an integer, the models are trained every refit windows.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB