Update monai-registration-notebook and fix bug

Closes #373 (closed).

• The bug was due to using torch.nn.MSELoss instead of torch.nn.functional.mse_loss
• The structure of the training plan was updated following Sergen's other examples:
  • remove __init__
  • add init_model, init_optimizer, init_dependencies
  • fix where necessary self.* -> self.model().*