DyC-CLIP: Dynamic context-aware multi-modal prompt learning for zero-shot anomaly detection | Synapse