Add Libra to LLM post-training repositories

[kaiwenKevinn]

Summary

• Add Libra to the RL & LLM Fine-Tuning Repositories section.
• Describe Libra as a resource-aware framework for asynchronous agentic RL post-training.
• Link the open-source repository and accompanying paper.

Libra provides disaggregated training and rollout pools, heterogeneous vLLM workers, dynamic resource planning, elastic execution, and asynchronous GRPO workflows for agentic language models.

• Repository: https://github.com/NetX-lab/Libra
• Paper: https://arxiv.org/abs/2606.03077
• License: MIT