We have a type safety issue in BindingDataObjectCodecTreeNode, where we incorrectly capture choice/case child relationships, harking back to the days before MDSAL-45, which introduced ChoiceIn marker interface.

The workaround is to go through intermediate raw BindingDataObjectCodecTreeNode, where the DataObject requirement is squashed down.

What we really need to do is separate these constructs, so that have an explicit intermediate BindingChoiceInCodecTreeNode and/or allow performing an immediate step across child nodes.