Fix get_feature_names_out raising AttributeError on fitted estimators

[LukeTheoJohnson]

Problem

Calling get_feature_names_out() on a fitted PCA, MCA, or FAMD estimator raises:

AttributeError: 'PCA' object has no attribute 'n_components_'

import pandas as pd, numpy as np, prince

X = pd.DataFrame(np.arange(40, dtype=float).reshape(10, 4), columns=list("abcd"))
pca = prince.PCA(n_components=3).fit(X)
pca.get_feature_names_out()  # AttributeError

This also breaks Pipeline.get_feature_names_out() and ColumnTransformer, which delegates to this method, so a pipeline containing a prince estimator fails the same way.

Cause

get_feature_names_out returns np.arange(self.n_components_), but there's nothing in fit that assigns n_components_. It just exists as a class-level annotation, so the type checker stays green while the attribute is missing at runtime.

Fix

fit now sets self.n_components_ = len(self.svd_.s) in PCA.fit and MCA.fit (FAMD inherits). Works in the paths I traced (including the rank capped CA/MCA SVD where fewer components are kept than requested) and matches the width of the transform output. Happy to adjust if there's a case I've missed.

Tests

One regression test per estimator, asserting len(get_feature_names_out()) == transform(X).shape[1]. All three raise the AttributeError on master and pass with this change. ruff check, ruff format --check, and ty check prince are clean.

[MaxHalford]

Cheers! I'll take a look in a couple of weeks after my holidays.