Detecting steps in spatial genetic data: which diversity measures are best?