From Sora What We Can See: A Survey of Text-to-Video Generation