This PR significantly improves the model card for [Vector Quantization using Gaussian Variational Autoencoder](https://huggingface.co/papers/2512.06609) by:
- Adding the `pipeline_tag: image-to-image` to the metadata for better discoverability and potential inference widget activation.
- Updating the paper link to the official Hugging Face Papers page.
- Including a link to the dedicated project page.
- Adding a concise model description based on the paper's abstract.
- Providing detailed sample usage code snippets (for both VQ-VAE and Gaussian VAE inference) directly from the GitHub repository, making it easier for users to get started.
- Adding the BibTeX citation.
Please review and merge if everything looks good.
