`OneVsOneClassifier` does not accept custom input types

Describe the bug

Due to the additional validation in #6626, OneVsOneClassifier cannot be used with custom types that work like arrays but cannot be converted to arrays. In my case I am the maintainer of scikit-fda, a functional data library that attempts to be compatible with scikit-learn. The metaestimators OneVsRestClassifier and OneVsOneClassifier should be trivially compatible with our data. Currently OneVsRestClassifier works fine, while OneVsOneClassifier doesn't.

I think that scikit-learn has sometimes very aggressive validation that makes it difficult to extend it for custom objects, as in this case.

Steps/Code to Reproduce

I have no example using only scikit-learn, as you need custom types and classifiers / transformers to trigger it
The following code is adapted from https://fda.readthedocs.io/en/latest/auto_tutorial/plot_skfda_sklearn.html#multiclass-and-multioutput-classification-utilities
Note that the code works if you replace OneVsOneClassifier with OneVsRestClassifier, as the later does not have that aggressive validation.

import skfda
import skfda.preprocessing.dim_reduction.variable_selection as vs
from sklearn.model_selection import train_test_split
from sklearn.svm import SVC
from sklearn.pipeline import Pipeline
from sklearn.multiclass import OneVsOneClassifier

X, y = skfda.datasets.fetch_phoneme(return_X_y=True)

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)

pipeline = Pipeline([
    ("dim_reduction", vs.RKHSVariableSelection(n_features_to_select=3)),
    ("classifier", SVC()),
])

multiclass = OneVsOneClassifier(pipeline)

multiclass.fit(X_train, y_train)
multiclass.score(X_test, y_test)

Expected Results

The result of the classification.

Actual Results

TypeError                                 Traceback (most recent call last)
TypeError: float() argument must be a string or a number, not 'FDataGrid'

The above exception was the direct cause of the following exception:

ValueError                                Traceback (most recent call last)
<ipython-input-2-66e12b97fe3e> in <module>
     17 multiclass = OneVsOneClassifier(pipeline)
     18 
---> 19 multiclass.fit(X_train, y_train)
     20 multiclass.score(X_test, y_test)

~/Programas/Utilidades/Lenguajes/miniconda3/envs/fda38/lib/python3.8/site-packages/sklearn/multiclass.py in fit(self, X, y)
    726         """
    727         # We need to validate the data because we do a safe_indexing later.
--> 728         X, y = self._validate_data(
    729             X, y, accept_sparse=["csr", "csc"], force_all_finite=False
    730         )

~/Programas/Utilidades/Lenguajes/miniconda3/envs/fda38/lib/python3.8/site-packages/sklearn/base.py in _validate_data(self, X, y, reset, validate_separately, **check_params)
    579                 y = check_array(y, **check_y_params)
    580             else:
--> 581                 X, y = check_X_y(X, y, **check_params)
    582             out = X, y
    583 

~/Programas/Utilidades/Lenguajes/miniconda3/envs/fda38/lib/python3.8/site-packages/sklearn/utils/validation.py in check_X_y(X, y, accept_sparse, accept_large_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, multi_output, ensure_min_samples, ensure_min_features, y_numeric, estimator)
    962         raise ValueError("y cannot be None")
    963 
--> 964     X = check_array(
    965         X,
    966         accept_sparse=accept_sparse,

~/Programas/Utilidades/Lenguajes/miniconda3/envs/fda38/lib/python3.8/site-packages/sklearn/utils/validation.py in check_array(array, accept_sparse, accept_large_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, ensure_min_samples, ensure_min_features, estimator)
    744                     array = array.astype(dtype, casting="unsafe", copy=False)
    745                 else:
--> 746                     array = np.asarray(array, order=order, dtype=dtype)
    747             except ComplexWarning as complex_warning:
    748                 raise ValueError(

ValueError: setting an array element with a sequence.

Versions

System:
    python: 3.8.0 | packaged by conda-forge | (default, Nov 22 2019, 19:11:38)  [GCC 7.3.0]
executable: /home/carlos/Programas/Utilidades/Lenguajes/miniconda3/envs/fda38/bin/python3.8
   machine: Linux-5.4.0-92-generic-x86_64-with-glibc2.10

Python dependencies:
          pip: 21.3.1
   setuptools: 60.5.0
      sklearn: 1.0.2
        numpy: 1.22.0
        scipy: 1.7.3
       Cython: 0.29.26
       pandas: 1.4.0
   matplotlib: 3.5.1
       joblib: 1.1.0
threadpoolctl: 3.0.0

Built with OpenMP: True

Start by reproducing the supplied scikit-fda example and read sklearn/multiclass.py around OneVsOneClassifier.fit and its _validate_data call. Compare its handling with OneVsRestClassifier, then verify that the same FDataGrid pipeline can fit and score with OneVsOneClassifier without the reported conversion error.