Benchmark library

Pick a problem before picking a model.

Each benchmark compares four implementation classes against explicit operating constraints, on a real public dataset. Methods that have not been run are shown as such rather than estimated.