Parallelize and chunk AutoML.predict

[lsorber]

Resolves #239.

[coveralls]

Coverage increased (+0.04%) to 75.544% when pulling 6fc0bb4 on lsorber:parallelize-automl-predict into 4530fc6 on automl:development.

[mfeurer]

Looks great, thank you very much. I think you forgot to pass the batch_size to the predict function of the actual model, though.

[lsorber]

All right, finally got around to passing the batch_size to the actual model. Does it look ok to you now?

[mfeurer]

Looks good, thank you very much.

[mfeurer]

Now we only need to rebase on the development branch. I can do this, but if I do it, would you still be the author of the commits? Otherwise, would you like to do this?

[lsorber]

I rebased the PR onto development, and squashed the commits to clean up the history. Should be good to go now, although it looks like the pyrfr build failing is preventing Travis from running the tests. Let me know if you need anything else.

[mfeurer]

Thanks a lot @lsorber for your contribution to auto-sklearn! Further contributions are highly appreciated ;)