agent simulation: enrich Turn metrics

[Bobronium]

Homogenizes Turn/Job/Run LLM/TTS/STT shapes

Turn: flat fields -> stt/llm/tts/conversation groups
      transcription_delay_ms -> stt.transcription_latency_ms

New:
  turns[].conversation.agent_yield_latency_ms      per barge-in; presence = barge-in
  turns[].conversation.turn_taking_score           0-1 per event
  turns[].conversation.{eot_misprediction,
    awkward_silence, unanswered}_count             0/1 per turn; replaces flags
  turns[].llm.tokens_per_second
  Conversation.response_latency_ms                 per-turn slot; unset on job/run
  TTS.heard_words, TTS.speech_ms                   speech_rate pooling stats

[theomonnom]

it's unclear what the flags could be, is it on purpose?

[Bobronium]

The field was intended to carry information about the turn such as "barge_in", "unanswered". "eot_misprediction", "awkward_silence". Comment was stale by the time of the PR.

Now with conversation message included (it contains dedicated fields for this), it's redundant and removed.

[changeset-bot]

🦋 Changeset detected

Latest commit: 2abd1c8

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
github.com/livekit/protocol	Patch
@livekit/protocol	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR