Association Feature Selection should only hide features visually
When reusing Association and making a selection within the class diagram (making a selection for a to-many association end), the features that should be hidden from the user are actually removed from the model and later re-added (using commands and then undo).
However, this causes problems with the impact model evaluations (see pull request #115). I.e., removing the feature One leaves StructureType with only one XOR child. This will result in an invalid configuration that is passed to Familiar, which causes a parse exception.
Instead, the features should only be hidden visually (i.e., no feature view should be created for those features), as is also done in other cases, such as reusing with re-exposed features and the different selection modes.
