Add Python 3.14 support

[pmachapman]

Adding Python 3.14 support is a bit more involved than 3.13. It involves (at least) the following breaking or not yet implemented upstream changes:

• Pyarrow does not yet support Python 3.14: https://arrow.apache.org/docs/python/install.html#python-compatibility
• Updating datasets to version 4.4.0 or higher (see https://github.com/huggingface/datasets/releases/tag/4.4.0)
  • This will likely break backwards compatibility with older versions of datasets loading datasets created with version 4.0 (see Replacement of "Sequence" feature with "List" breaks backward compatibility huggingface/datasets#7739)
• Updating to a future (>=5?) version of transformers
  • The dependency tokenizers uses Pyo3 which must be version 0.27 or higher (https://github.com/PyO3/pyo3/releases/tag/v0.27.0)
  • tokenizers at time of writing only uses Pyo3 0.26 (https://github.com/huggingface/tokenizers/releases/tag/v0.22.2)
  • This will change a few namespaces plus a few breaking changes exposed by unit tests

These are likely to break downstream dependencies...