M3V: Multi-modal Multi-view Context Embedding for Repair Operator Prediction