Support for sparse matrices in StackingClassifier with use_features_in_secondary=True

[FlorisHoogenboom]

When using StackingClassifier with use_features_in_secondary=True and using a sparse matrix as input one runs into a value error on calling fit: ValueError: all the input arrays must have same number of dimensions. This seems caused by the fact that the results of the meta estimator cannot be concatenated to the sparse matrix.

Would you consider adding support for sparse matrices with use_features_in_secondary set to True? If so, I would be happy to open a PR for this!

[rasbt]

Thanks for the note! Having support for thios would be awesome, and I would appreciate a PR!

I think this might also affect the StackingCVClassifier, StackingRegressor, and StackingCVRegressor. However, we can tackle them one by one, and starting with the StackingClassifier would be great

[FlorisHoogenboom]

You are probably right, I believe the logic for merging the meta estimator output and the original feature matrix is the same for all of those. I hope to have some time next week to have a look at those as well and will submit a PR for those seperately.

[rasbt]

Thanks for the PR the other day! After all this physical labor associated with moving I felt like coding today and already implemented it for the rest, in case you are wondering

[FlorisHoogenboom]

Awesome!