[feature request] Parse Input and Output Messages in Langchain Instrumentor

[Jgilhuly]

Is your feature request related to a problem? Please describe.
We don't currently parse out the actual response from langchain / langgraph as plain text in our instrumentor. We instead dump the whole llm response dictionary, which is difficult to read as a user.

Describe the solution you'd like
I recognize langchain and langgraph have some inconsistency in their output formats, but ideally we could add some conditional logic to at least parse out these plaintext values in the most common scenarios.

Additional context

[axiomofjoy]

Thanks @Jgilhuly, can you drop a screenshot of what you mean?