Added typing and structure to SdoTermSource.

The various fields of SdoTerm contain either identifiers (as strings), or instances of SdoTerm and the code hacks around this with a lot of obscure casting operations. Replaced the content with an object than contains either, or a sequence which contains either. This makes it possible to at least document the types that are taken by various methods. This also enables, making the code smarter, as most of the current functions systematically assume the worst.

• Added holding structures.
• Added unit-tests for these structures.
• Updated all the code that access theses fields, including the Jinga templates.
• Added typing information in many places.
• Removed test that was broken but shadowed by another test with the same name (thanks python).
• Added documentation where I figured out what is going on.
• Added some assertions on types in critical entry points.