Resource Management for Low-latency Cooperative Fine-tuning of Foundation Models at the Network Edge