Translate context documents in the benchmark and use combinations of languages

At the moment, all context documents are in English while questions are in English, German and French. It would be interesting to also (automatically) translate all context documents to German and French and to evaluate all combinations of context and question language to better understand the limitations of the different models in terms of multilingual capabilities.