Add new new metric: Comet

[cjlovering]

• Added dependency
• Added wiring for using it by default in a translation task
• (We should discuss the above -- i.e. is it desired?)
• Uses the same framework as SARI -- expects a doc_to_rawtext to be
  implemented

TODOs

• Test on a translation task. (Tentatively works, but I'm having memories issues on my laptop running the associated model or else some other problem is coming up. I will look into it further.)
• Determine if we want to run it by default on all translation tasks (as is setup right now)

[StellaAthena]

@rbawden, do you think we should use this as the default?

[rbawden]

I'm not really sure. BLEU is still the most used metric by far I'd say, so we should include results for it and it might be wise to keep it as default (despite all the criticisms). I would tentatively say COMET could be a secondary metric, but it could be worth checking with some other people to see what the consensus is.

[cjlovering]

Update: I think that computing COMET after the pipeline has been run would be cleaner. I will return to this in a few days.