post_fit_renderer for active learning ?

[remiadon]

Would be great to have a post_fit renderer, i.e a method called after each fit on a batch
to render stats like scoring

Would be instanciated by default to a print, but user could pass it as an arg, eg

python aflow.active --post-fit-render="lambda estimator, X, y: print(estimator.score(X, y))" ... 

or

python aflow.active --post-fit-render=mymodule.render_function ....