Text this: M3AE-Distill: An Efficient Distilled Model for Medical Vision–Language Downstream Tasks