add an example of how to classify images to the docs

[niedakh]

A simple case like sklearn's sift + vbow + multilabels, perhaps on mnist?

See:
https://ianlondon.github.io/blog/how-to-sift-opencv/