TODO:

  * Port over and improve clusterview and it's opengl stuff. It should use k-means point and cluster, as well as
    the math library from kmeans (add a test to dist). Additionally, the point and cluster of the clusterview should
    inherit from k-means point and cluster so that it can use the algorithm correctly. All other aspects of point and
    cluster in the clusterview program can be kept.
  * Extract and improve the overall structure of clusterview so that globals are not used unless absolutely necessary.
  * Turn kmeans into a python package and import it here.
  * Remove old clusterview buttons, keep the status window side and saving feature. Add new buttons for weighted and
    unweighted clustering.
  * Add a property to the point called weight, which is set when you click edit point. It will give a popup that
    allows you to specify a point weight.
  * Use kmeans to do the stuff.
  * Weighted cluster mean is rendered as a hollow circle and unweighted k-means mean is a x.