feat: add skeletoken

[stephantul]

This PR replaces a lot of tokenization code with simple calls and validation from skeletoken.

This has the following effects:

1. the code is much simpler
2. from now on, every tokenizer is a greedy tokenizer, with optional lower casing (set to True by default)

This improves scores greatly on ModernBERT (which is a cased BPE tokenizer).

This PR also removes Python 3.9 support, since skeletoken does not support Python 3.9 because Pydantic does not have great support for unions in 3.9.

[codecov]

Codecov Report

❌ Patch coverage is 89.81481% with 11 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
model2vec/tokenizer/tokenizer.py	91.35%	7 Missing ⚠️
model2vec/distill/distillation.py	80.95%	4 Missing ⚠️

Files with missing lines	Coverage Δ
model2vec/distill/inference.py	97.53% <100.00%> (+0.03%)	⬆️
model2vec/tokenizer/__init__.py	100.00% <100.00%> (ø)
model2vec/tokenizer/datamodels.py	100.00% <100.00%> (ø)
model2vec/utils.py	90.38% <100.00%> (-0.36%)	⬇️
model2vec/distill/distillation.py	87.50% <80.95%> (+1.29%)	⬆️
model2vec/tokenizer/tokenizer.py	94.24% <91.35%> (+3.71%)	⬆️

🚀 New features to boost your workflow:

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