Add missing return type hints to CharBPETokenizer.from_file, train, train_from_iterator

[RudrenduPaul]

What this PR does

Adds missing return type hints to three methods in
bindings/python/py_src/tokenizers/implementations/char_level_bpe.py:

• from_file(vocab_filename: str, merges_filename: str, **kwargs) -> "CharBPETokenizer"
• train(...) -> None
• train_from_iterator(...) -> None

This follows the same typing-consistency pattern as #2211, which added
the equivalent missing return hints to BaseTokenizer.save,
save_model, and to_str. CharBPETokenizer (a BaseTokenizer
subclass) had the identical gap in its own methods: every argument was
already annotated, but the return type was never added.

Why these types are correct

• from_file is a @staticmethod that returns CharBPETokenizer(vocab, merges, **kwargs),
  i.e. a new instance of the class — hence -> "CharBPETokenizer", matching the quoted
  self-referential forward-reference convention already used elsewhere in this codebase
  (e.g. Tokenizer.from_file -> "Tokenizer" and BPE.from_file -> "BPE" in the generated
  .pyi stubs).
• train and train_from_iterator both call self._tokenizer.train(...) /
  self._tokenizer.train_from_iterator(...) without a return statement, so they implicitly
  return None — hence -> None.

Scope note

Checked gh pr list --repo huggingface/tokenizers --search "char_level_bpe" before starting;
no open or closed PR currently touches this file.

No behaviour change

Type-hint-only change on method signatures. No logic, docstrings, or other files were modified.

Testing

• ruff check bindings/python/py_src/tokenizers/implementations/char_level_bpe.py — all checks passed.
• ruff format --check bindings/python/py_src/tokenizers/implementations/char_level_bpe.py — passes, file already formatted.
• python3 -c "import ast; ast.parse(...)" on the modified file — passes.

Note: Claude Code was used to assist in drafting this change. All changes were reviewed by the submitter.