This PR adds a comprehensive model card to the repository for the Legal$\Delta$ model.
The model card includes:
- A link to the paper: [Legal$\Delta$: Enhancing Legal Reasoning in LLMs via Reinforcement Learning with Chain-of-Thought Guided Information Gain](https://huggingface.co/papers/2508.12281).
- A link to the official GitHub repository: https://github.com/NEUIR/LegalDelta.
- A detailed description of the model and its core features, including the paper's abstract.
- A sample usage section to help users get started with inference, noting the LoRA checkpoint nature.
- Relevant metadata, including the license (`apache-2.0`), `library_name` (`transformers`), and `pipeline_tag` (`text-generation`), which will improve discoverability on the Hugging Face Hub.
Please review the changes.
