CATSE: A Context-Aware Framework for Causal Target Sound Extraction