Future Annotation Improvements

[PGijsbers]

Some notes for later as I go along with writing type hints, on things that I can't tackle because they require a newer Python version or a larger refactoring than I want to do right now (I want some test coverage first).

• The file formats that are stored by string should be StrEnum when we move to 3.11+
• Introduce namedtuples for "complex" return types, like the loaded benchmarks
• Enums for options of the Encoder of the DataUtils (or see if we can somehow remove that altogether. It's there now so we don't have a dependency on sklearn in the framework virtual environment)

[PGijsbers]

Other remarks, not typing related:

• not self.delegate should be a none check for encoder