feat: Multilingual retrieval loader

[Samoed]

Checklist

• Run tests locally to make sure nothing is broken using make test.
• Run the formatter to format the code using make lint.

I've added a multilingual retrieval loader with the following structure:

• {config}-corpus
• {config}-queries
  ...

However, currently, there are no datasets that fit this structure. Previously, when I uploaded datasets, I used corpus-config, but more other datasets start with config. These datasets have variations that make them incompatible with this loader for direct use, so this is more for the future

[orionw]

I like it! Seems like it should make it easier. Have you checked that it works with some of the multilingual datasets? If you have then it seems good to merge to me.

[Samoed]

I tested this on CodeSearchCCRetrieval, which uses corpus-{config}. It worked when I swapped the name in the corpus for the test. However, I couldn't find datasets where the loader could be changed due to different naming schemes. This PR won't break anything; it's an attempt to standardize for future use

[KennethEnevoldsen]

Looks good might be nice to add a push_to_hub method that pushes a dataset in a standard format (would make it easy to convert datasets)