Gender Bias in Text-to-Video Generation Models: A case study of Sora