Fix drop_unknown_references for null foreign keys

[R-Palazzo]

Environment Details

SDV main

Error Description

Currently, the drop_unknown_references doesn't work as expected for foreign keys with null values.
This bug arises because having null values inside the foreign keys is valid, and so metadata.validate_data(data) no longer crashes.

To fix it, drop_unknown_references should check for null foreign keys and update the data accordingly, also according to its parameter drop_missing_values.

Steps to reproduce

The following snippet code currently fails while it should work

parent = pd.DataFrame(data={
        'id': [0, 1, 2, 3, 4],
        'A': [True, True, False, True, False],
        'B': [0.434, 0.312, 0.212, 0.339, 0.491]
    })

    child = pd.DataFrame(data={
        'parent_id': [0, 1, 2, 2, 5],
        'C': ['Yes', 'No', 'Maye', 'No', 'No']
    })

 data =  {
        'parent': parent,
        'child': child
}
data['child'].loc[4, 'parent_id'] = np.nan

cleaned_data = drop_unknown_references(metadata, data)
metadata.validate_data(cleaned_data)

pd.testing.assert_frame_equal(cleaned_data['parent'], data['parent'])
pd.testing.assert_frame_equal(cleaned_data['child'], data['child'].iloc[:4])
assert len(cleaned_data['child']) == 4