In the era of the rise against statistical significance: what have we (not) learnt from millions of scientific papers with P-values