Get some benchmarks and comparisons

[AtheMathmo]

It would be good to get some benchmarks for rusty-machine to see how we compare to other common libraries. Perhaps the most obvious would be scikit learn/mlpack. Of course there are some issues in a direct comparison as it is likely that these libraries offer more stable solutions. The numbers will be good to have regardless.

We should also be wary of the use (and lack of) LAPACK/BLAS. It would be enough to just note explicitly when this is in use and not.

It would be great to have comparisons for a number of models, showing time to train and time to predict in each (with as similar parameter settings as we can achieve across the board). Showing this in a table (or tables) within a BENCHMARKS.md file in the project root would probably be ideal. The machine details used to produce any values should be clearly stated as well.

Any contribution to this effect would be greatly appreciated!