[FIX] Select Rows: Removing Unused Values for Discrete Variables in Sparse Data

[nikicc]

Issue

When remove unused features is checked in Select Rows and some discrete variable comes from sparse data, Orange crashes with AttributeError: ravel not found error.

The problem is in the remove_unused_values method (Orange/preprocess/remove.py) which cannot handle sparse matrices.

Issue: https://sentry.io/biolab/orange3/issues/269413491/

Description of changes

• Implement nanunique function. nanunique returns unique values without missing (np.nan) values and works on sparse and dense matrices.
• Fix remove_unused_values to use nanunique and hance support sparse discrete columns.

Includes

• Code changes
• Tests
• Documentation

[codecov-io]

Codecov Report

Merging #2452 into master will decrease coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2452      +/-   ##
==========================================
- Coverage   74.51%   74.49%   -0.02%     
==========================================
  Files         321      321              
  Lines       56056    56055       -1     
==========================================
- Hits        41769    41760       -9     
- Misses      14287    14295       +8