Changeset:
* `scprep.plot.histogram` supports 0 and negative data with `log='x'` and `log=True`
* `scprep.filter.filter_values` (`library_size`, `gene_set_expression`) supports `cutoff=(low, high)` and `percentile=(low, high)`
* variable genes calculation is more stable using a rolling median (no longer uses loess)
* added `scprep.measure.gene_capture_count`
* added `scprep.stats.t_statistic` and `scprep.stats.rank_sum_statistic`
* `scprep.stats.differential_expression` supports `measure='ttest'` and `measure='ranksum'`