Reorganize the "Classifying PanSTARRS with (un)supervised learning" entry

[cmurray-astro]

Add example to its own folder with its own requirements.txt file

[jegpeek]

great tutorial! small comments:

• How is the ~1.9 billion sources consistent with the 2,902,054,648 PS1 sources? Warps vs. stacks?
• apply principal component analysis, t-distributed stochastic neighbor embedding (tSNE) -> apply principal component analysis (PCA), t-distributed stochastic neighbor embedding (tSNE)
• This might be for later, but maybe it's worth trying it for just three classes? UNSURE is kind of strange thing to try to classify. In particular this comment: Based on the above confusion matrix, the SGD classifier performs very well at identifying stars (~94% true positive) whereas it does not perform as well at distinguishing between the galaxy, QSO and unsure classes. seems not to be true when you look at just the 9 boxes that don't include UNSURE, right?
• DOI to the paper as in https://github.com/spacetelescope/hellouniverse/pull/5
• assistant astronomer / deputy title comment as in https://github.com/spacetelescope/hellouniverse/pull/5