Voice mode: auto-submit transcription on Space release

When using voice mode (hold Space to record), releasing Space transcribes the audio but does not submit it. The user still has to press Enter manually.

This defeats the purpose of a hands-free voice workflow. The expected behavior is that releasing Space should automatically submit the transcribed text, similar to how voice assistants work.

At minimum, this should be a configurable option (e.g., a setting like voiceAutoSubmit: true or a dedicated keybinding).

Current workflow: Hold Space → speak → release → read transcription → press Enter
Expected workflow: Hold Space → speak → release → done