Null models for species co-occurrence patterns: assessing bias and minimum iteration number for the sequential swap