add logging callback for llm completion events

[willsmanley]

Now consumers can log all LLM completion requests like this:

def will_log_completion_event(cht_ctx, collected_text, tool_calls, interrupted):
   my_logger(cht_ctx, collected_text, tool_calls, interrupted)

assistant = VoiceAssistant(
      vad=silero.VAD.load(),
      stt=deepgram.STT(),
      llm=openai.LLM(),
      tts=openai.TTS()
      chat_ctx=initial_ctx,
      will_log_completion_event=will_log_completion_event
)

It gets the updated cht_ctx from will_synthesize_agent_reply.

In general I would like to build out more of this type of audit trailing. Other things I am thinking about are first class support for testing (see hamming.ai or similar).