Add MarianNMT

[xhluca]

See Marian: https://huggingface.co/transformers/model_doc/marian.html
See helsinki-nlp's models: https://huggingface.co/Helsinki-NLP

We'd need

• Add option to load the marian architecture at initialization (e.g. dlt.TranslationModel("marian"))
• Add an option to find all of the languages (and code) available for a certain variant trained using marian, e.g. dlt.utils.available_languages("opus-en-romance")
• An option to leverage autocomplete such as dlt.lang.opus.en_romance.ENGLISH, but the options would be limited to only what's available with the variance (i.e. romance)
• TBD

[xhluca]

Could we auto-gen the languages for OPUS as well? we could use it for:

# 
dlt.lang.opus.romance.ENGLISH
dlt.lang.opus.north_eu.GERMAN

And get the available languages using:

dlt.utils.available_languages("opus", source="en", target="romance")

Wonder if above should return a dictionary {"source": [...], "target": [...]}, a tuple of (source, target)? Or we would need to specify what to return: dlt.utils.available_languages("opus", source="en", target="romance", return_type="source")?

[xhluca]

Actually, this might make more sense:

dlt.lang.romance.ENGLISH

dlt.utils.available_language("romance")

A lot more concise and logical.

[xhluca]

Already available in easynmt, which has good support for multiple models