feat(telemetry): track successful voice starts

[johnmatthewtennant]

Summary

Berd cannot currently distinguish whether voice conversations are used. This adds the consent-gated berd_voice_conversation_started counter after native voice startup succeeds. The event has no attributes and collects no session, transcript, voice selection, or failure details.

Uploads opt into the strict berd-otlp-logs-v2 contract that accepts this counter.

Notes

berd-otlp-logs-v2 must be deployed in squareup/berd-monitoring before releasing this client version. End-to-end gateway verification requires that deployment.

[morgmart]

🤖 Automated code review

APPROVE. Fresh static review completed for the exact three-dot comparison d67756c...f43d7c2. The consent-gated voice-start counter is emitted only after native startup and microphone reconciliation succeed, carries no event attributes, and is covered for successful and failed startup. The schema-version update and exact wire/event literals are pinned by tests. Final self-check covered the changed voice-start and telemetry-upload flows; accessibility and localization (no changed UI or user-facing copy); navigation and consent guards; async failure, never-completes, lifecycle, and race behavior; test honesty; project/design-system rules; duplicate overlap; and severity evidence/user effect. No publishable findings survived. Supplied GitHub evidence was inspected: all nine check runs at the exact head SHA completed successfully, while the combined commit status is pending with no legacy statuses; required checks still govern merge readiness.

Deterministic publication result: 0 blocking and 0 non-blocking finding(s) publishable; 0 duplicate(s) suppressed.

Pending checks: 1 check(s) are not complete.

This approval reflects the completed code review only; merge readiness remains governed by the repository's required checks.