The label classifier will attempt to use labels that it trained
against in `models/label_tokenizer.json`.

However, labels often change. Instead of retraining the whole data
set, you may want the label classifier to ignore certain labels and
attempt to find the next best prediction.

In `scripts/config.yml`, you can add labels that will be skipped. For
example, if you want to skip over `group::package`, you can add an
entry to the YAML file.

Fixes https://gitlab.com/gitlab-org/ml-ops/tanuki-stan/-/issues/7