feat: add util for tokenizer pad id

[stephantul]

The padding token in our classifier module defaulted to [PAD], which works for our own models (models that have bert-base-uncased as a base), but not for e.g., e5 based models. This required that users manually look up their padding token and set it.

Because we now use skeletoken, we actually store the model's pad token, so we can just look it up. I introduced a new utility that tries to guess the pad token and which will show a warning if we default to 0.

pad_token is still an accepted argument, but is now set to None by default, so this is a backwards compatible change. For users that didn't fill in this argument, it will still work, because we guess that [PAD] can be a padding token.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
model2vec/train/base.py	97.97% <100.00%> (+0.08%)	⬆️
model2vec/train/utils.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Pringled]

LGTM