Replace MLX GPU inference with CoreML bundles so transcription continues while the host app is backgrounded. Adds model download and warm-up, vendored Qwen3Speech, and updates onboarding, settings, and copy for the ~1.6 GB CoreML package (iOS 18+).
- Custom Keyboard Extension with push-to-talk UI - iOS 26 SpeechAnalyzer + DictationTranscriber (iOS 18 SF fallback) - OpenAI-compatible LLM client (4 built-in providers + custom) - 3-page onboarding flow + provider config UI - App Group shared storage for cross-process config - 8s LLM timeout with raw-transcript fallback - App Store privacy manifests for both targets - SwiftLint + XcodeGen + GitHub Actions CI - Unit tests (4/4 passing)