A Closer Look at Knowledge Distillation in Spiking Neural Network Training