More informative error when calling transform of unfitted metric learner

[bellet]

We should throw an informative error/exception when one attempts to use an unfitted transformer to transform data (or any other method which requires the metric learner to be fitted). See below what happens in metric-learn versus sklearn. We can probably borrow some code from sklearn to handle this similarly.

Example for metric-learn:

>>> from metric_learn import NCA
>>> from sklearn.datasets import make_classification
>>> X, y = make_classification()
>>> nca = NCA()
>>> nca.transform(X)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/aurelien/Documents/research/github/metric-learn/metric_learn/base_metric.py", line 244, in transform
    preprocessor=self.preprocessor_,
AttributeError: 'NCA' object has no attribute 'preprocessor_'

Example for sklearn:

>>> from sklearn.decomposition import PCA
>>> from sklearn.datasets import make_classification
>>> X, y = make_classification()
>>> pca = PCA()
>>> pca.transform(X)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/aurelien/anaconda3/lib/python3.7/site-packages/sklearn/decomposition/base.py", line 125, in transform
    check_is_fitted(self, ['mean_', 'components_'], all_or_any=all)
  File "/home/aurelien/anaconda3/lib/python3.7/site-packages/sklearn/utils/validation.py", line 914, in check_is_fitted
    raise NotFittedError(msg % {'name': type(estimator).__name__})
sklearn.exceptions.NotFittedError: This PCA instance is not fitted yet. Call 'fit' with appropriate arguments before using this method.

[bellet]

Done in #267