KANDINSKY Patterns: A Swiss-Knife for the Study of Explainable AI