Inferring the impact of linguistic boundaries on population differentiation: application to the Afro-Asiatic–Indo-European case