Exploring Parameter-Efficient Fine-Tuning of a Large-Scale Pre-Trained Model for scRNA-seq Cell Type Annotation