Cross-Space Synergy: A Unified Framework for Multimodal Emotion Recognition in Conversation