Upgrade machine.py production docker container to python 3.11

[mshannon-sil]

Upgrading the production docker container to python 3.11. I also changed some of the tests I wrote because when I at first updated all of the poetry packages, a few of them stopped passing. However, I've since reverted most of the packages back to their original versions and only upgraded packages that were necessary to upgrade, but I kept the changes to the tests regardless for maximum compatibility.

---

This change is 

[johnml1135]

tests/translation/huggingface/test_hugging_face_nmt_model_trainer.py line 145 at r1 (raw file):

            training_args,
            corpus,
            src_lang="en_XX",

Why would the test without the _XX fail on newer versions of python? We should support language codes that don't include a script code.

[johnml1135]

Reviewed 3 of 4 files at r1, all commit messages.
Reviewable status: 3 of 4 files reviewed, 1 unresolved discussion (waiting on @ddaspit and @mshannon-sil)

[mshannon-sil]

Reviewable status: 3 of 4 files reviewed, 1 unresolved discussion (waiting on @ddaspit and @johnml1135)

---

tests/translation/huggingface/test_hugging_face_nmt_model_trainer.py line 145 at r1 (raw file):

Previously, johnml1135 (John Lambert) wrote…

Why would the test without the _XX fail on newer versions of python? We should support language codes that don't include a script code.

I don't think it was the fact that it was a newer version of python that was making it fail; rather, I had updated transformers without needing to, and the updated version caught that I was using the wrong language codes for this tokenizer. The tokenizer that the tiny-random-nllb model uses is actually the slow MBartTokenizer rather than NllbTokenizerFast, maybe because MBartTokenizer takes a lot less space for testing. That tokenizer does include script codes in all of its default language codes. I'm not sure why it wasn't failing earlier when I was using the wrong language codes for this tokenizer, but I guess that one of the most recent updates just fixed it to work properly and fail if it sees an unrecognized language code. So I don't think we need to make any further changes to these tests.

[johnml1135]

Reviewed 1 of 4 files at r1.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @ddaspit)