Implementing probabilistic KDE error bounds - Week 2

This week has been more about trying to solve a problem rather than implementing it. I couldn't get this improvement to work as it should, so I asked for help to my mentor Ryan. Finally, after a week of trials and thanks to his help, we're making some progress :)

Estimations seem to be getting to a reasonably good point and might fully work as expected soon. This sounds like a good time to start implementing the actual API for handling the new KDE parameters.

I have also been exploring how trees are implemented in mlpack, this is going to be useful for the next improvement we'll be working on.

"An algorithm must be seen to be believed." - Donald Knuth.