This PR improves the model card for Mega-ASR by:
- Adding links to the official paper ([2605.19833](https://huggingface.co/papers/2605.19833)), project page, and GitHub repository.
- Adding the `datasets` metadata for the training set used ([Voices-in-the-Wild-2M](https://huggingface.co/datasets/zhifeixie/Voices-in-the-Wild-2M)).
- Including a Python sample usage section based on the official repository.
- Organizing the content for better readability.
