Refactor README to Improve Onboarding and Technical Clarity

[chuangw6]

The project's main README.md lacked an immediate definition of the core product, missed explicit local setup instructions, and suffered from vertical bloat due to a large inline code snippet that disrupted reading flow for newcomers.

Proposed Solutions

• Add a clear elevator pitch describing Open-RL as a lightweight backend for advanced LLM post-training.
• Highlight its core capabilities.
• Provide explicit local setup instructions using existing Makefile targets.
• Have a structured Guide section by spliting links strictly into SFT vs. RL paths to prevent confusion for newcomers.
• Rewrite "Why Tinker?" to clearly contrast the benefits for AI Researchers vs. Platform Engineers.
• Encapsulate the lengthy GRPO Python snippet within an HTML <details> tag to keep the reading flow clean while preserving technical depth.