NotInitializedError inherits from NotFittedError (#860)

This ensures that a user can catch the sklearn `NotFittedError`, just as
they can with regular sklearn estimators. As discussed in [1].

[1] #859

Author: timokau

skorch/exceptions.py

@@ -1,11 +1,13 @@
 """Contains skorch-specific exceptions and warnings."""
 
+from sklearn.exceptions import NotFittedError
+
 
 class SkorchException(BaseException):
     """Base skorch exception."""
 
 
-class NotInitializedError(SkorchException):
+class NotInitializedError(SkorchException, NotFittedError):
     """Module is not initialized, please call the ``.initialize``
     method or train the model by calling ``.fit(...)``.