RaP: Redundancy-aware Video-language Pre-training for Text-Video Retrieval