[RAG & v0.2.1] build + eval + optimization + doc + notebook

[liyin2015]

• [x] build RAG
• [x] eval RAG
• [x] improve the evaluation guide in general and add more code tutorials/evaluation/eval_nlg.py
• [x] G-eval, no reference, relevance, consistency, fluency, coherence
• [x] First draft of llm evaluation guide (to do: editing the evaluation.rst)
• [x] finalize the metrics
• [ ] hotpotqa dataset
• [ ] doc
• [ ] notebook
• [ ] Optimization
• [ ] optimization doc
• [ ] optimization botebook
• [ ] setup notebook to colab process (keep in sync)