Apoc

New features
* `TableRowClassifier.train()` can consume table with `NaN` elements, which will be ignored while training
* `TableRowClassifier.train()` can consume `ground_truth` list entries `==0` which will be ignored while training
* Added a new predefined feature set `v070`

Backwards compatibility breaking changes
* `TableRowClassifier.predict()` has no parameter `return_numpy` anymore and returns a numpy array.
* Input features and results of `TableRowClassifier` `train()` and `predict()` are now always 1-dimensional. 2D arrays with shape `(1, n)` are no longer allowed.

Thanks to kevinyamauchi for feedback and support!

New features:
* Measurements and `APOC_ObjectClassification_CLUSTER_ID` are stored in `classifier._data` after object classification

New features
* `TableRowClassifier` for object classification beased on custom measurement tables (Thanks to kevinyamauchi for implementing this)

Bug fix
* Type error in error message when a wrong type of classifier is loaded

New features
* `apoc.list_available_object_classification_features()`
* `classifier.statistics()`

New features
* `apoc.train_classifier_from_image_folders` allows training from image/label-mask folder pairs

Bug fixes
* Removed debug traces to stdout