v1.7.1

patch for 1.7.0

• Tray sync on CLI model switch: Tray menu now reflects model changes made via speak-to-ai model set

Features

• Whisper model management: Choose between base, small, medium, and large-v3 quantized models — switch via system tray or CLI without restarting the daemon
• Model hot-reload: Seamless engine swap on model change — closes old engine, downloads if needed, creates new engine while mutex-protecting recording state
• CLI model commands: model list (offline), model set <id>, model delete <id> — manage models from terminal or scripts
• Download cancellation: Re-click a downloading model in tray to cancel; context-based cancellation propagates through the download pipeline
• Tray model submenu: Visual indicators for active model (●), available (○), downloading (⏳), and cancel-on-reclick (✖)

Based on #76 , #77

---