This PR adds a comprehensive dataset card for the TACO dataset.
It includes:
- A link to the paper ([TACO: Taming Diffusion for in-the-wild Video Amodal Completion](https://huggingface.co/papers/2503.12049))
- The `video-to-video` task category and relevant tags in the metadata
- Links to the project page (`https://jason-aplp.github.io/TACO/`) and GitHub repository (`https://github.com/JasonAplp/TACO`)
- The `cc-by-nc-4.0` license
- A detailed description of the dataset, its structure, and how to use it for inference and training, including installation steps and command examples.
