sklearn/utils/tests/test_pprint.py is too brittle to API changes

Although I was fine making this test file depend on the definitions of various estimators, #13470 has made it clear that this can be a nuisance to developers when they add a parameter to the estimators that happen to be used in those tests (or even change a default value). I think if we replace:

from sklearn.linear_model import LogisticRegression

with something like

# Some example constructors excerpted to test pprinting

class LogisticRegression(BaseEstimator):
    """Logistic Regression (aka logit, MaxEnt) classifier.
    """

    def __init__(self, penalty='l2', dual=False, tol=1e-4, C=1.0,
                 fit_intercept=True, intercept_scaling=1, class_weight=None,
                 random_state=None, solver='warn', max_iter=100,
                 multi_class='warn', verbose=0, warm_start=False, n_jobs=None,
                 l1_ratio=None):
        pass

and do similar for other classes, then we will get most of the effect of the tests (checking that standard use cases look acceptable) without being brittle to changes in parameters of those estimators.

[NicolasHug]

Feel free to give it a shot @hermidalc (or anyone else).

Else I’ll do it

[NicolasHug]

Closing since it was addressed in #13529