[FIX] Silhouette Plot: Fixes

[ales-erjavec]

Issue

• When the number of clusters was the same as the number of instances the widget crashed with an ValueError: Number of labels is {num} Valid values are 2 to n_samples - 1 (inclusive) error from scikit-learn.
• If the cluster column contained missing values the widget would crash with an assertion error (assert Ck[0] >= 0 and Ck[-1] < len(values)) in SilhouettePlot.setScores

Description of changes

• Check (and show an error message) if the number of clusters is the same as the number of instances
• Omit instances with missing cluster assignment from the display (and the "Selected Data" output) and show a warning when this happens. When the "Add silhouette scores" is selected, output missing values for scores corresponding to instances with unknown cluster assignment.

Includes

• Code changes
• Tests
• Documentation

[codecov-io]

Current coverage is 88.87% (diff: 100%)

Merging #1747 into master will not change coverage

@@             master      #1747   diff @@
==========================================
  Files            82         82          
  Lines          8819       8819          
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
  Hits           7838       7838          
  Misses          981        981          
  Partials          0          0          

Powered by Codecov. Last update 909a119...c7efeed