Provide more clarity about which level of text representation is used

We should add a way for users to set the `target_text_representation_level` from inside the wizard, and also indicate how to change it in the main documentation guide since it is not intuitive to set it otherwise. In this wizard, this (and other model-specific questions) should take place after the dataset loop so that we can determine whether there is an option (e.g. all datasets need to have phones available if we want to set `target_text_representation_level` to `phones`).