Preserving Semantic and Temporal Consistency for Unpaired Video-to-Video Translation