One hot encoder

[danthedaniel]

What does this PR do?

• Adds the autosklearn one hot encoder
• Adds two new config dictionaries with sklearn operators that support sparse matrices
• Refactors tests into a tests directory, broken into multiple files

Where should the reviewer start?

• tpot/one_hot_encoder.py
• tests/*.py
• TPOTBase._setup_config

I have not made any changes to the docs yet.

All credit to @slcott who originally PR'd these changes.

[coveralls]

Coverage increased (+1.0%) to 88.081% when pulling 39afc35 on teaearlgraycold:one_hot_encoder into 305701c on rhiever:development.

[coveralls]

Coverage increased (+1.0%) to 88.081% when pulling 74399b2 on teaearlgraycold:one_hot_encoder into 305701c on rhiever:development.

[coveralls]

Coverage increased (+1.0%) to 88.081% when pulling 74399b2 on teaearlgraycold:one_hot_encoder into 305701c on rhiever:development.

[coveralls]

Coverage increased (+1.0%) to 88.081% when pulling a0583e5 on teaearlgraycold:one_hot_encoder into 305701c on rhiever:development.

[rhiever]

Looks like this needs to be rebased with your latest repo reorganization PR.

[coveralls]

Coverage increased (+1.1%) to 88.678% when pulling 373097c on teaearlgraycold:one_hot_encoder into 995f8cc on rhiever:development.

[coveralls]

Coverage increased (+1.09%) to 88.65% when pulling aae9952 on teaearlgraycold:one_hot_encoder into 995f8cc on rhiever:development.

[weixuanfu]

The PR looks very nice. Only one thing about unit tests: OneHotEncoder has a separated code file one_hot_encoder_tests.py for the unit tests of this built-in operator. Should we also make separated files of unit tests for the other built-in operators (StackingEstimator and ZeroCount)? Or just combine unit tests for all built-in operators into a single file?

[rhiever]

@weixuanfu2016, we can do that separately in another PR. Merging this one before any other big changes happen and it needs to be rebased for the 20th time. :-)