[FIX] impute: Remove class vars from input data for ReplaceUnknownsModel

[ales-erjavec]

Issue

impute.Model does not work on input dataset with class_vars since gh-3925 due to error raised by backmapping.

import Orange
from Orange.modelling import ConstantLearner
from Orange.preprocess import impute

data = Orange.data.Table("brown-selected")
impvar = impute.Model(ConstantLearner())(data, data.domain[0])
impvar.compute_value(data)

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/aleserjavec/Documents/workspace/orange3/Orange/preprocess/impute.py", line 175, in __call__
    predicted = self.model(data[mask])
  File "/Users/aleserjavec/Documents/workspace/orange3/Orange/base.py", line 437, in __call__
    backmappers, n_values = self.get_backmappers(data)
  File "/Users/aleserjavec/Documents/workspace/orange3/Orange/base.py", line 248, in get_backmappers
    f"Model for '{modelclass.name}' "
Orange.data.table.DomainTransformationError: Model for 'alpha 0' cannot predict 'function'

Description of changes

Remove class vars from input data for ReplaceUnknownsModel

Includes

• Code changes
• Tests
• Documentation

[codecov]

Codecov Report

Merging #5083 (d061f53) into master (29e992f) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #5083   +/-   ##
=======================================
  Coverage   84.74%   84.74%           
=======================================
  Files         286      286           
  Lines       60036    60037    +1     
=======================================
+ Hits        50876    50878    +2     
+ Misses       9160     9159    -1