This PR enhances the CSEMOTIONS dataset card by:
- Adding crucial metadata: `task_categories` (text-to-speech), `library_name` (datasets), `license` (apache-2.0), `language` (zh for Mandarin Chinese), and relevant `tags` (speech, emotional-speech, voice-cloning, mandarin).
- Including direct links to the associated paper ([Marco-Voice Technical Report](https://huggingface.co/papers/2508.02038)) and the project's GitHub repository ([AIDC-AI/Marco-Voice](https://github.com/AIDC-AI/Marco-Voice)) at the top of the dataset card for improved discoverability.
- Correcting the number of speakers in the "Dataset Summary" for consistency with the paper and "Dataset Details".
- Restructuring the card to place metadata at the very top, followed directly by Markdown content, in line with Hugging Face Hub best practices.
