Extracting robust trends in species’ distributions from unstructured opportunistic data: a comparison of methods