Kwai-Klear/KlearReasoner-CodeSub-15K · Improve dataset card: Add paper link, task category, tags, code link, and data format

This PR significantly enhances the dataset card by adding crucial information for users:

Linking to the associated paper: Klear-Reasoner: Advancing Reasoning Capability via Gradient-Preserving Clipping Policy Optimization.
Adding task_categories: text-generation to the metadata for better categorization.
Including relevant tags such as reasoning, math, code, and reinforcement-learning for improved discoverability on the Hugging Face Hub.
Providing a direct link to the main Klear-Reasoner GitHub repository.
Adding a sample usage example to illustrate the dataset's structure and format for easier adoption.