feat(android): continuous hands-free voice in the active thread #48

Request

Add opt-in continuous voice to the Android app, similar to Dash: start once, detect speech boundaries, submit final utterances into the current thread, then listen again.

Triage

Current mobile has single-shot voice-input infrastructure; reuse it rather than inventing a second transcription/authentication path. Related broader app-level assistant RFC: #38. This issue is thread-scoped, not a cross-thread autonomous assistant. Spoken replies should be optional; full two-way voice versus continuous dictation remains an explicit design decision, not a claimed requirement.

Acceptance

• Visible listening/transcribing/error states and explicit stop/mute.
• One final utterance results in at most one submission; preserve manual drafts.
• Bind capture to exact environment/thread; stop on navigation, background, disconnect and permission loss.
• Respect current queue/steer delivery policy while a turn is running.
• Reuse existing transcription capability; no credentials in app or silent paid provider fallback.
• Focused lifecycle tests; real Android microphone/audio acceptance reported separately.

Fork-local implementation first. No upstream PR or device install without separate authorization.