Lack of tag in classifiers causes problems for scorers using `predict_proba`

[Heptasphere]

Describe the bug

A scorer created through scikit-learn which uses predict_proba can't be used with classifiers from mlxtend.classifiers

Classifiers are tagged by default as regressors for scikit-learn with versions above 1.6.
This interferes with scikit-learn's scorers using predict_proba.

Similar problem as rasbt/machine-learning-book#205.

Steps/Code to Reproduce

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.metrics import roc_auc_score, make_scorer
from mlxtend.classifier import MultiLayerPerceptron

X, y = load_breast_cancer(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)

roc_auc_scorer = make_scorer(
    roc_auc_score,
    response_method="predict_proba",
    multi_class="ovo",
)

model = MultiLayerPerceptron()
model.fit(X_train, y_train)

result = roc_auc_scorer(model)
print(result)

Expected Results

A float between 0 and 1 is printed, showing the performance of the model.

Actual Results

ValueError: MultiLayerPerceptron should either be a classifier to be used with response_method=predict_proba or the response_method should be 'predict'. Got a regressor with response_method=predict_proba instead.

Versions

MLxtend 0.23.4
Linux-6.8.0-79-generic-x86_64-with-glibc2.35
Python 3.12.11 (main, Jun 4 2025, 08:56:18) [GCC 11.4.0]
Scikit-learn 1.7.1
NumPy 2.3.2
SciPy 1.16.1

Additional Context

A similar error occurs with EnsembleVoteClassifier, LogisticRegression SoftmaxRegression and StackingClassifier.

StackingCVClassifier had an entirely different error based on fit_params but it would likely suffer the same bug.

Adaline, OneRClassifier and Perceptron don't have a predict_proba method so it is not reproductible this way but they likely share this tag problem.

Partial Solution

The same solution of switching ClassifierMixin and BaseEstimator as in rasbt/machine-learning-book#205 works for EnsembleVoteClassifier and likely works for OneRClassifier.
Other classifiers use different mixin that don't seem to implement tags.

[Heptasphere]

Directly adding a method __sklearn__tags to classifiers with the proper tags without changing the order of inheritance also seems to work. More explanations at https://scikit-learn.org/stable/developers/develop.html#estimator-tags.
Not tested for class other than EnsembleVoteClassifier but it should solve this problem for every classifier. 

from sklearn.utils._tags import ClassifierTags

class EnsembleVoteClassifier(BaseEstimator, ClassifierMixin):

    ...

    def __sklearn_tags__(self):
            # Next line might not be necessary for classes not inheriting from BaseEstimator
            tags = super().__sklearn_tags__() 
            tags.estimator_type = "classifier"
            tags.classifier_tags = ClassifierTags()
            tags.target_tags.required = True
            return tags

It may be a good idea to add the other tags each classifier is supposed to have too.