"Cuvier had preceded Lamarck in specifying the kinds and degrees of variation, which his own observations and critical judgment of the reports of others led him to admit."