Add menu bar quick selectors for dictation microphone, language, and model

[SeoFood]

Problem

Changing microphones, spoken languages, or transcription models currently requires leaving the immediate dictation flow for settings. The menu bar exposes the active provider/model label and dictation actions, but does not provide equivalent direct selectors.

Proposed behavior

Add compact, keyboard-accessible selectors to the menu bar for microphone, language/language set, and transcription provider/model. Reuse existing selection services and preferences. Show the effective selection, including workflow overrides, rather than suggesting that a global default always controls the next recording.

Start with the menu bar. Any later indicator controls should use the same state and actions.

Acceptance and verification

• Show selected and unavailable/disconnected devices, with clear system-default behavior.
• Preserve existing language-selection semantics and model compatibility restrictions.
• Show missing/download-required models without starting an undisclosed download.
• Define whether changes during recording are disabled or apply to the next session; never silently interrupt capture.
• Keep selections synchronized with Settings and workflow overrides.
• Verify keyboard navigation, VoiceOver, full-screen apps, and focus preservation for subsequent insertion.

Priority: P2. Estimated effort: medium.

Source: MenuBarView.swift.
Related: #1214 concerns the System Audio Recorder surface; this issue is for system-wide dictation quick selection.