Text this: Parameter-Efficient Adaptation of Large Vision—Language Models for Video Memorability Prediction