------------------
* eli5.sklearn.InvertableHashingVectorizer and
eli5.sklearn.FeatureUnhasher allow to recover feature names for
pipelines which use HashingVectorizer or FeatureHasher;
* added support for scikit-learn linear regression models (ElasticNet,
Lars, Lasso, LinearRegression, LinearSVR, Ridge, SGDRegressor);
* doc and vec arguments are swapped in explain_prediction function;
vec can now be omitted if an example is already vectorized;
* fixed issue with dense feature vectors;
* all class_names arguments are renamed to target_names;
* feature name guessing is fixed for scikit-learn ensemble estimators;
* testing improvements.