[feat] explore and document how to access guardrails validation outcome metadata when using langchain

[dtam]

Description see how we can better expose the validation outcome metadata when using langchain, openai and the guardrails server openai endpoint

Why is this needed want to know about error spans, validation errors, etc

Implementation details

End result