Flesh out tools for using confidence when drafting

[mshannon-sil]

More features need to be added to SILNLP to support the use of confidence scores in the translate step. Currently, a .confidences.tsv file is created when translating which keeps track of the confidence scores for each verse and each token in each verse. We're missing the following functionality:

1. Inserting verse confidence scores into the draft itself, so that the info is all in one place for the translator to look at.
2. The ability to obtain a chapter confidence score during the translation step. Currently this can only be done through running diff_predictions.py with --chapter, but diff_predictions requires the presence of test data.
3. The ability to obtain a book confidence score during the translation step. Currently, this can only be done through running test.py with --score-by-book, which requires test data.

[mshannon-sil]

After further discussions, this is the revised list of features to be added for now:

1. The ability to obtain a chapter confidence score during the translation step. Currently this can only be done through running diff_predictions.py with --chapter, but diff_predictions requires the presence of test data.
2. The ability to obtain a book confidence score during the translation step. Currently, this can only be done through running test.py with --score-by-book, which requires test data.
3. A flag for test.py, translate.py, and experiment.py to save confidence scores, rather than saving confidence by default, since we're still determining how best to use the confidence scores.

[ddaspit]

@mshannon-sil Has this been completed?

[mshannon-sil]

Yes, looks like I linked to it in the body of the PR, but not under the "Development" header on the sidebar, so it didn't automatically close.