The Y-axis scale is not starting from 0 (zero) and the lowest scale shows up on the Y-axis of the result is set based on the lowest count value, which makes the impression that the lowest count sample has hardly any counts, which is not true and is misleading. If the count number is very large or wide spread between the filenames, we should be able to choose log-scale or spread the bar in 2 parts (by providing a high/low separation criteria, for instance). The choice of linear or log scale should be provided as a optional.

GH workflows have to be populated and tests must be replaced with latest framework (test -> tests)

https://github.com/tercen/count_operator/tree/0.10.0

Usage of this operator written in README is not very clear, "input projection" in particular. For example, to count the cell number (event number) per sample (filename), this operator requires factors in Y-axis (e.g. value) and column (e.g. filename).

Output: Y-axis: counts, column: filename/sample

It is the best to automatically choose Bar as a default projection, instead of a dot.