Update Corrective RAG demo with small fixes

[AndresHG]

I have been playing with Corrective RAG demo in LangGraph repository and gave found some small issues and typos. Here is the list of them:

• OpenAI model GPT-4o-mini is cheaper and supports structured output as opposite to gpt-3.5-turbo used along the script
• Added a small section to load dotenv and quick-setup for LangSmith
• Some prints at the end of the script were broken or not showing the expected output
• Typo: "ALL DOCUMENTS ARE NOT RELATED TO THE QUESTION..."; actually this is triggered if at least 1 of them is not relevant, so the message is wrong.

[AndresHG]

Hello @eyurtsev , dotenv is something I added at the end, but not really relevant, just a helpful thing I considered, but I can quickly delete that. What about the other typos and warnings fixed? Running the notebook with GPT 3.5 will generate a warning or a braking error.

[eyurtsev]

Ah let me check -- I may have been fooled by GitHub rich display

[hesreallyhim]

@eyurtsev is this still open? happy to add a review if so, but don't want to waste time if not.

[sydney-runkle]

We're going to do a docs refactor here quite soon, going to close for now, but can ping you for feedback when we return to this resource. Thanks!