Abductive explanations of classifiers under constraints: Complexity and properties