Add images data classification w/ SVM and squeeze out a lot of performance

@mikolajr Motivation:

• classical images analysis also very good
• connects later to NNs getting even more performance

See MNIST SVM solutions from Kaggle / rankings page.

Alternatively: celebrities faces dataset classification w/ SVM from scikit-learn (also in VanderPlass book)

In script about pipelines or details on classifiers e.g. as a summary exercise (dep. on if tunning needs some insights).