K-means clustering for boosting

# Feature
## Desired Behavior / Functionality
The boosting module is, right now, setting the number of points that should be used. A different approach could be to apply a clustering algorithm (motivated by the "why train more paper" as the points seem to cluster in two groups).

## What Needs to Be Done
Implement Kmeans clustering on the scores.

## How Can It Be Tested
Run the attacks.