Finish up ROCAUC

ROCAUC needs a few things to be finalized:

• decision function bug (binary classifier)
• documentation
• add to gallery
• simplify example in docstring and switch to YB data instead of sklearn.datasets
• add multiclass and binary fixtures from conftest.py (see ClassificationReport
• remove X, yb, and ym prefering to use above fixtures
• remove load_binary_data and load_multiclass_data to use above fixtures

Note the bug is as follows:

======================================================================
ERROR: Test ROCAUC with a binary classifier
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/benjamin/Repos/ddl/yellowbrick/tests/test_classifier/test_rocauc.py", line 110, in test_binary_rocauc
    s = visualizer.score(X_test, y_test)
  File "/Users/benjamin/Repos/ddl/yellowbrick/yellowbrick/classifier/rocauc.py", line 171, in score
    self.fpr[i], self.tpr[i], _ = roc_curve(y, y_pred[:,i], pos_label=c)
IndexError: too many indices for array

[rebeccabilbro]

Looks like setting the random_state for LinearSVC and RandomForestClassifier did the trick!

[rebeccabilbro]

Closed with #540