TALC: Time-Aligned Captions for Multi-Scene Text-to-Video Generation