speechd: Separate out voice variants
So the user can first choose the language (most often already correctly set by the locale) then the variant proposed by the server for the language.
So the user can first choose the language (most often already correctly set by the locale) then the variant proposed by the server for the language.