Add pydantic and json_schema to German extractor

[empiriker]

This is to add pydantic classes to the German extractor.

In some cases, pydantic fields are defined as lists of strings where probably they could be just a string. I decided to accept for now this looser schema to avoid refactoring even more. This was already quite a bit more painful than I anticipated.

[xxyzz]

Thanks for your contribution!

How's the performance compare to the previous code?

[empiriker]

Without pydantic:

2023-12-06 10:12:39,031 INFO:   ... 20000/20000 pages (100.0%) processed, 00:00:00 remaining
2023-12-06 10:12:39,048 INFO: Reprocessing wiktionary complete
         776069 function calls (776037 primitive calls) in 98.246 seconds

With pydantic:

2023-12-06 10:08:30,500 INFO:   ... 20000/20000 pages (100.0%) processed, 00:00:00 remaining
2023-12-06 10:08:30,515 INFO: Reprocessing wiktionary complete
         577305 function calls (577273 primitive calls) in 98.502 seconds

These are the processing times of running each version one time on my local machine. Apparently pydantic comes at a cost of 0.26% of performance.

Obviously a more rigorous test setup might give more insights but to me this seems encouraging enough. What do you think?

[xxyzz]

That's pretty insignificant. We probably don't need to worry about it.