This PR adds a comprehensive model card for the ModernVBERT model, linking it to the paper [ModernVBERT: Towards Smaller Visual Document Retrievers](https://huggingface.co/papers/2510.01149).
It includes the license, `library_name` (transformers), and `pipeline_tag` (visual-document-retrieval) in the metadata for better discoverability and integration on the Hub. The content provides a concise description of the model, an architecture overview, and links to relevant resources including the GitHub repository, Hugging Face organization, blog post, and a Google Colab tutorial for usage.
Please review and merge if everything looks good.
