Logging metrics for agents

[fniedtner]

support following metrics for parity with 0.2:

• Agent Conversation History
• No of Call to the LLM
• LLM Latency
• Request and Response Token count
• Agent Interaction Count
• LLM Request and Response Payload

[victordibia]

I think these are addressed with model_usage and support for otel