[iOS] Add native text-to-speech playback for transcripts and generated results

[SeoFood]

Context

TypeWhisper can create transcripts, notes, email drafts, and other generated results, but iPhone and iPad users cannot ask the app to read that content aloud. Native text-to-speech would improve hands-free review and accessibility while keeping text on device.

The first version should use Apple's on-device speech synthesis unless a documented requirement cannot be met without an additional provider.

Proposed scope

• Add read-aloud entry points for the current result and History detail.
• Support transcripts and generated documents such as notes and email drafts.
• Provide native play, pause, resume, stop, and progress controls.
• Choose an appropriate voice automatically from the output language, with a simple user override where needed.
• Define predictable audio-session behavior when recording, playback, phone calls, other interruptions, wired audio, or Bluetooth audio is active.
• Keep generated speech ephemeral for the first version; audio export can be evaluated separately.
• Keep transcript and document text local when using the native Apple speech baseline.

Acceptance criteria

• Users can start and stop read-aloud playback from a current result and a History entry.
• Play, pause, resume, stop, elapsed progress, and completed state remain synchronized with actual playback.
• Automatic voice selection works for German, English, French, Italian, and Spanish output, with a usable fallback when the preferred voice is unavailable.
• Starting a recording stops or safely pauses speech playback; starting playback never leaves the recording audio session active.
• Speaker, wired, and available Bluetooth routes behave consistently, and route changes do not leave stale playback UI.
• Phone calls and common system interruptions pause or stop playback predictably and allow a documented recovery path.
• VoiceOver does not duplicate spoken content unexpectedly, and all playback controls have clear accessibility labels.
• iPhone and iPad layouts work in portrait and landscape at supported Dynamic Type sizes.
• Tests cover playback state transitions, cancellation, interruption recovery, language/voice fallback, and audio-session ownership.

Design decisions before implementation

• Confirm whether background or lock-screen playback belongs in the first release.
• Decide whether voice and rate are global settings or per-playback overrides.
• Validate that AVSpeechSynthesizer covers the accepted first-release behavior before considering network providers.