[ENH] K-means: clusters can be inferred for new data

[markotoplak]

Issue

@borondics had a so big data set that k-means was too slow, so he tried doing it on a sample and then modelling the clustering with a classifier. But for k-means this should not be needed because we could use the means/medoids to "predict" the cluster directly.

With one additional line we could make the Cluster useful with Apply domain.

Why don't we already do it? All the machinery is already there, in ClusteringModel, which does seems unused though. Does

Includes

• Code changes
• Tests
• Documentation

[codecov]

Codecov Report

Attention: Patch coverage is 95.00000% with 1 line in your changes missing coverage. Please review.

Please upload report for BASE (master@66928eb). Learn more about missing BASE report.
Report is 23 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #7010   +/-   ##
=========================================
  Coverage          ?   88.72%           
=========================================
  Files             ?      332           
  Lines             ?    73444           
  Branches          ?        0           
=========================================
  Hits              ?    65164           
  Misses            ?     8280           
  Partials          ?        0           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[markotoplak]

@janezd, I finally finished __eq__ and __hash__. To make it work, I needed to change some attributes into read-only properties (which also made me rewrite a test that relied on changing them).