Evaluating the Effectiveness of Large Language Models in Representing and Understanding Movement Trajectories