How can we categorise ‘nationality’ and ‘second generation’ in surveys without (re)producing stigmatisation?