Privacy Auditing of Multi-domain Graph Pre-trained Model under Membership Inference Attacks