This PR significantly enhances the model card for `3DGraphLLM` by:
* Adding `pipeline_tag: image-text-to-text` to categorize the model's functionality for better discoverability on the Hub.
* Adding `library_name: transformers` to indicate compatibility with the Hugging Face Transformers library for programmatic use.
* Including additional relevant `tags` such as `3d-scene-understanding`, `scene-graph`, `multimodal`, `vlm`, and `llama`.
* Adding the full paper abstract to provide a detailed overview of the model directly on the card.
* Updating the paper link to the official Hugging Face Papers page: [3DGraphLLM: Combining Semantic Graphs and Large Language Models for 3D Scene Understanding](https://huggingface.co/papers/2412.18450).
* Including a direct link to the official GitHub repository for code access.
* Incorporating the performance benchmark table from the GitHub README for quick assessment of the model's capabilities.
* Adding a "Usage" section that directs u…
