Sparse Tuning Enhances Plasticity in PTM-based Continual Learning