Langroid

* [Neo4jChatAgent](https://github.com/langroid/langroid/tree/main/langroid/agent/special/neo4j) to chat with a neo4j knowledge-graph. The agent translates user queries to Cypher queries and the queries are
run against an existing, specific knowledge-graph. See [example script](https://github.com/langroid/langroid/tree/main/examples/kg-chat)
* Support for .doc file parsing (in addition to .docx)

Update with recent changes to azure, aiohttp dependency

Minor fix: Agent.base - _get_one_tool_message - handle case when request is not a str

User conveniences:
* Ignore bad OPENAI_API_KEY when user is specifying a non-OpenAI model
* DocChatAgentConfig - no need to specify RelevanceExtractorAgentConfig.llm: if not specified, it will inherit LLM config from
DocChatAgentConfig. But of course one could specify these to use two different LLMs.

RelevanceExtractorAgent: Handle segment-marking for text that has lost all/most punctuation, making it impossible to identify sentences.

LanceDocChatAgent: Handle empty query in query plan