Revisiting Active Learning in the Era of Vision Foundation Models