Backdoor Attack in Prompt-Based Continual Learning