With the upcoming major release, should we consider changing the default of the normalize parameter in the various (PC) scores from True to False? This would better align with the scikit-learn API, potentially reducing confusion for users familiar with that framework. Do you have strong opinions on this @slevang ?