This PR adds a model card for SwimBird, including:
- Metadata for the `image-text-to-text` pipeline tag.
- Links to the [research paper](https://huggingface.co/papers/2602.06040), [project page](https://accio-lab.github.io/SwimBird), and [GitHub repository](https://github.com/Accio-Lab/SwimBird).
- A summary of the model's core contribution: query-adaptive switching between text-only, vision-only, and interleaved reasoning modes.
- BibTeX citation information.
