The bug of reading contents from md file in demo code

[HaiyangPeng]

After running the code PageIndex/examples/agentic_vectorless_rag_demo.py, the following case will occur:

1. when the file is md, LLM will infer multiple page_num, e.g., 53,57 and the _get_md_page_content may return contents from line 53 to line 57. This is not correct for md file, so I fix this bug by changing the agent prompt to use ONLY the exact line_num values from the
   structure, limiting the tool description of get_page_content and using precise line extraction in _get_md_page_content.

I will provide a PR for fixing this bug