feat(voice): add OpenAI speech-to-text

[johnmatthewtennant]

Summary

Adds OpenAI realtime speech-to-text as a Voice Conversation input backend.

• Adds a dedicated OpenAI STT API-key field in Voice settings.
• Stores the STT key under its own keychain account and never reads Goose provider credentials or the OpenAI TTS key.
• Waits for the realtime transcription session to be ready before accepting microphone audio.
• Preserves committed transcript delivery while discarding stale turns across mute boundaries.
• Keeps STT and TTS readiness and keychain errors independent.

Configuration

Enter an OpenAI speech-to-text key in Settings → Voice → Speech input after selecting OpenAI. This credential is separate from Goose and from the OpenAI TTS key.

Testing

• just check
• pnpm vitest run src/features/voice-conversation/hooks/useOpenAiVoiceSetup.test.tsx src/features/voice-conversation/ui/VoiceSettings.test.tsx src/features/voice-conversation/lib/voiceSetupReadiness.test.ts
• targeted OpenAI transcription Rust tests
• just tauri-check
• just clippy