feat(examples): add arXiv papers knowledge base demo

[moshierming]

Summary

This PR adds a new unofficial sample demonstrating how to build a searchable knowledge base from arXiv papers using LightRAG's graph-enhanced retrieval.

What this example does

1. Fetches paper metadata via the free arXiv API (no API key required)
2. Inserts papers into LightRAG for graph-indexed storage
3. Demonstrates Local / Global / Hybrid query modes on academic content

Usage

# Using OpenAI
python examples/unofficial-sample/lightrag_arxiv_papers_demo.py \
    --ids 2410.05779 1706.03762 2005.11401

# Using local Ollama (free, no API key)
python examples/unofficial-sample/lightrag_arxiv_papers_demo.py \
    --ids 2410.05779 1706.03762 \
    --llm-model qwen2.5:7b \
    --embed-model nomic-embed-text \
    --ollama

Key features

• No extra dependencies beyond lightrag-hku — arXiv fetching uses stdlib urllib
• Retry logic with exponential backoff for environments with unstable arXiv API access
• Supports both OpenAI and local Ollama backends
• Cleanly separates paper ingestion from query demo, easy to extend

Testing

Syntax-checked with python -m py_compile. Full integration test requires LLM credentials.

[danielaskdd]

Thanks for the interest! The demo looks great. Please go ahead with the submission—I'd be happy to consider merging it.