Adopt MVVM structure for voice-to-text feature

[fbraza]

Summary

Adopt the MVVM architecture for the voice-to-text feature and organize the project structure accordingly.

Requirements

• Restructure the lib/ tree to follow a feature-first layout with MVVM layers.
  • core/ for shared models, services, utilities.
  • features/voice_to_text/ containing view/, viewmodel/, models/, widgets/, and optional feature services.
• Introduce a VoiceToTextViewModel that exposes observable state (transcript, active word index, timer, waveform data, recording state) and command methods (start/pause/stop recording, update transcript/waveform).
• Ensure view widgets depend only on ViewModel outputs and forward user interactions to ViewModel commands.
• Inject dependencies (audio capture, speech recognition, storage) via constructors using a DI mechanism (Provider, get_it, or equivalent).
• Add unit tests for the ViewModel covering state transitions and service interactions.

Acceptance Criteria

• Project folders reflect the MVVM structure and feature-first organization.
• VoiceToTextViewModel exists with defined state and command API, decoupled from UI widgets.
• Voice-to-text widgets consume ViewModel state via dependency injection without direct service calls.
• ViewModel tests validate core logic paths with mocked services.
• Documentation (README or architecture doc) outlines the MVVM pattern for contributors.