CategoricalImputer fails when fill_value is already a category #1113

CategoricalImputer raises when fill_value is already present in a pandas categorical dtype, instead of filling the missing values.

import pandas as pd
from feature_engine.imputation import CategoricalImputer

X = pd.DataFrame({
    "region": pd.Categorical(
        ["east", None, "west"],
        categories=["unknown", "west", "east"],
    )
})
CategoricalImputer(fill_value="unknown", variables=["region"]).fit_transform(X)

This raises ValueError: new categories must not include old categories: {'unknown'}. I expected ['east', 'unknown', 'west'], with the categorical dtype preserved. It also fails with an ordered categorical; the equivalent object column works.

Reproduced on current main (d866312), Python 3.12, pandas 2.3.3 and scikit-learn 1.7.2. This is related to #1095, but #1096 changes RareLabelEncoder, not this imputer's separate add_categories call.

I'm happy to contribute the fix and regression tests, and I'm planning to work on this.