implement Elkan's algorithm for k-means

[alxndrkalinin]

Elkan's algorithm is an improved version of usual kMeans (called Lloyd's algorithm) that uses the triangular inequality to avoid many distance calculations when assigning points to clusters.

• [ ] read how algorithm works
• [ ] create new service kMeansElkan and inject it into ClusterKMeans
• [ ] service should provide method step that implements Elkan's algorithm
• [ ] create a list of kmeans algorithms similar to initialization methods
• [ ] display dropdown allowing user to choose the specific algorithm similar to choosing init methods
• [ ] test that algorithms works using using Neuroimaging study of 27 AD, MCI, NC subjects dataset and compare it's convergence with normal kmeans
• [ ] submit pull request to SOCR/SOCRAT:dev

Note: for algorithm implementation you can look at scikit-learn Cython code as an example and use ClusterMetrics service to calculate distances