Enlighten-Anything: When Segment Anything Model Meets Low-Light Image Enhancement