[sort] Investigate the performance of hana::sort

As outlined in https://github.com/edouarda/brigand/issues/118#issuecomment-189834748, `hana::sort` seems to perform best in the case of a reversed list, which is the only benchmark that we have. We should investigate this and probably change the benchmark.

Start with the discussion in brigand issue 118, comment 189834748, and inspect the existing hana::sort benchmark for reversed lists. Compare performance across relevant input orderings and update the benchmark if the current case is misleading; done means the performance behavior is investigated and the benchmark reflects the findings.

Written by the indexing model from the issue text.