Explain R2 score the first time we introduce it

In the first course of Regularization in linear model, you mention the R2, but it is never defined anywhere else before.

I think in general you switch often between different types of metrics (mean square error, score, R2) which makes it hard to comprehend for someone new to this field.

It also relates to my previous comment earlier today.

I created another topic about switching between different metrics and made this one about explaining R2 score the first time we introduce it.

Done in https://github.com/INRIA/scikit-learn-mooc/pull/325