feat: For a ML task, include some baselines

[sylvaincom]

Is your feature request related to a problem? Please describe.

Data scientists often rush to use relatively complex models on a new data set they discover, without trying out some simple baselines first.

Describe the solution you'd like

Add some baseline models.

Skore should be able to detect that the user has not tried some baselines first and make some recommendations. This could be LLM-assisted?

[glemaitre]

There is stuff before to do LLM-assisted stuff: https://skrub-data.org/stable/reference/generated/skrub.tabular_learner.html

[sylvaincom]

Yes! I was also thinking about the dummy estimators of scikit-learn? At least just to compare the score of my model, to see if it beats very simple baselines