Better selection of main input/output

[joshreini1]

This change better extracts the content from main_input and main_output to both display in the TruLens UI and use for helper methods such as .on_input() and .on_output(). Particularly, this change looks for content following the OpenAI pattern, e.g.

Inputs such as:

'{
     "model": "gpt-3.5-turbo",
     "messages": [{"role": "user", "content": "Say this is a test!"}],
     "temperature": 0.7
   }'

And outputs such as:

{
    "id": "chatcmpl-abc123",
    "object": "chat.completion",
    "created": 1677858242,
    "model": "gpt-3.5-turbo-0613",
    "usage": {
        "prompt_tokens": 13,
        "completion_tokens": 7,
        "total_tokens": 20
    },
    "choices": [
        {
            "message": {
                "role": "assistant",
                "content": "\n\nThis is a test!"
            },
            "logprobs": null,
            "finish_reason": "stop",
            "index": 0
        }
    ]
}

Notice in both cases, the text we actually care about is located as the value for the key, content. This pattern is widely used and adopted by frameworks such as Pinecone's Canopy and middleware such as LiteLLM.

See an example of this change below:

Example 1: Pinecone Canopy

Before:
This warning in notebook:

Unsure what the main input string is for the call to chat with args [[UserMessage(role=<Role.USER: 'user'>, content='How can you get started with Pinecone and TruLens?')]].
Unsure what the main output string is for the call to chat with return type <class 'canopy.models.api_models.ChatResponse'>.

And

After:

Example 2: OpenAI

Before:

After:

[piotrm0]

Perhaps this can fallback to just returning str(all_args) maybe with a prefix "Could not determine main input/output of {str(all_args)}." ?