Add Bayesian generative classification with KDE decontamination algorithm

[Gabriel-p]

Investigate the possibility of using KDEs obtained from field stars and the cluster region to assign AAAAAAAAAAAAAMPs.

The general idea is that through the field KDE we estimate the star's probability of belonging to the field, P(B), and with the cluster region KDE the probability of belonging to the mix of cluster + field stars, P(A+B) or P(C).

The challenge is to derive P(A) from these, ie: the probability of belonging to the cluster.

Article about this:
https://jakevdp.github.io/PythonDataScienceHandbook/05.13-kernel-density-estimation.html

---

Add 21/10/19

This issue is closely related to the new membership algorithm we are developing.