Efficient SVDD sampling with approximation guarantees for the decision boundary