Introduce cross-validation earlier in module 1 and related changes

[lesteve]

• 1. add a new notebook between the notebook 1 and 2 to introduce the model evaluation. It should discuss train_test_split and introduce cross_validate. This would be a notebook called something like "Evaluate your first model".
• 2. We would move some content from https://inria.github.io/scikit-learn-mooc/python_scripts/02_numerical_pipeline_scaling.html#model-evaluation-using-cross-validation but some discussions can be reduced since we don't use a Pipeline anymore.
• 3. If necessary, add a small section only speaking about the Pipeline inside cross_validate (I guess when this happens later in the "Preprocessing for numerical features" notebook
• 4. "Exercise M1.03" -> check if we should add an additional exercise that uses cross-validation. Consensus: let's not add an exercise that may be too simple and distract from the main ideas.

[ArturoAmorQ]

I would add

• 5. Add discussion/plots on score distributions
• 6. Add a figure similar to this one illustrating cross-validation. Good enough for now.

[lesteve]

Add discussion/plots on score distributions

If it is consise enough why not, I would have thought that module 2 was more the right place for a longer discussion (see #416). I guess we can quickly show here that the score depends on the train-test split which motivates the introduction of cross-validation

Add this figure illustrating cross-validation

The figure you mention has things about hyperparameter tuning so we probably don't want to include this exact figure. We already have a cross-validation figure: https://inria.github.io/scikit-learn-mooc/python_scripts/02_numerical_pipeline_scaling.html#model-evaluation-using-cross-validation. If you think it can be improved (which is quite likely) you can look at the script generating the figure: https://github.com/INRIA/scikit-learn-mooc/blob/master/figures/plot_cross_validation_diagram.py

[ArturoAmorQ]

My idea / proposition for points 5. and 6. Notice that the distribution plot matches the example.

[lesteve]

I think most of this has been tackled.