Add variant distribution constraint

[0xbe7a]

This PR adds a VariantDistributionConstraint which checks if the distribution of values in a column falls within the specified minimum and maximum bounds.

[codecov]

Codecov Report

Merging #136 (c12b4c9) into main (c7f5a4b) will not change coverage.
The diff coverage is 60.00%.

@@           Coverage Diff           @@
##             main     #136   +/-   ##
=======================================
  Coverage   36.44%   36.44%           
=======================================
  Files          15       15           
  Lines        1723     1723           
=======================================
  Hits          628      628           
  Misses       1095     1095           

Impacted Files	Coverage Δ
src/datajudge/requirements.py	50.51% <57.14%> (ø)
src/datajudge/constraints/uniques.py	33.05% <100.00%> (ø)

[YYYasin19]

I like this a lot! 🚀

To give a concrete example that might highlight how this could be used:
In our project, we have a timestamp column representing the birthdate of a person. When reading in data, this value gets messed up sometimes. This results in a lot of outliers (e.g. people are suddenly born in 1762) but also a lot of misinterpreted data which causes the values to all be around ~1970.

With this test, we could quantize all timestamps to their respective year resulting in a categorical column with values in [1900, 1901, ..., 2023] and then check that each category only takes up at most 5% of the column -- preventing accumulations.

[kklein]

Really great work!

[kklein]

What would you think of a relative violation tolerance parameter? E.g. it could say:

A test succeeds iff
#observations outside of the specified ranges / #observations <= tolerance_parameter

I don't consider it a must - we've simply faren well with tolerances historically.

[0xbe7a]

If 'A' is expected to have a target share ranging from 5% to 15%, but its actual share is 16%, would you consider the 16% to be a violation of the target range or merely 1% above the upper limit?

[0xbe7a]

I added this feature

[kklein]

Looks great - thanks a bunch! :)