StoryTTS: A Highly Expressive Text-to-Speech Dataset with Rich Textual Expressiveness Annotations