From Speech and PPG to EDA: Stress Detection Based on Cross-Modal Fine-Tuning of Foundation Models | Synapse