[feature] Support OpenAI client attribute extraction by patching client

[spike-spiegel-21]

Please have a look at this PR before merging: #1551

This enables tracing LlmAttributes for OpenAI client by patching it in Observer.

Example:

client = OpenAI(api_key="<your-openai-api-key>")

@observe(model="gpt-4o-mini", type="llm", llm_provider="openai", client=client)
def generate_response(input: str) -> str:
    response = client.chat.completions.create(
        model="gpt-4o-mini", 
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": input},
        ],
        temperature=0.7,
        tools=tools,
        tool_choice="auto"
    )
    return response

Note: Sample test in test_open_client_trace.py.

[vercel]

@spike-spiegel-21 is attempting to deploy a commit to the Confident AI Team on Vercel.

A member of the Team first needs to authorize it.