The caption model selector now sits directly under the dictation model in
the STT tab's Local Whisper card, where people look for it, and stays in the
General tab's caption section as well; both use one shared component. The
first selector is renamed to 'Dictation model' so the two are told apart.
