fix(tts): add versioned extended Latin grapheme support

[rohan-patnaik]

What does this PR do ?

Adds opt-in extended-Latin support to IPA dictionary parsing and word tokenization while preserving existing vocabularies by default.

Collection: TTS

Changelog

• Add latin_charset_version=2 for Latin Extended-A/B and Latin Extended Additional.
• Apply the version consistently in English and locale-agnostic tokenizers and IpaG2p.
• Keep version 1 as the default and warn once for unsupported Unicode-letter entries.

Usage

Set latin_charset_version=2 when the tokenizer should admit extended Latin graphemes. Omitting it preserves legacy behavior and token IDs.

GitHub Actions CI

93 broader tokenizer/G2P tests, 45 focused tests, and pre-commit pass locally. External CI may require /ok to test cfcdfec08b0ba7d437051450a967289c4fc8458c.

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install?
  • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• New Feature
• Bugfix
• Documentation

Additional Information

Related to #16083. Legacy mode retains the shipped German dictionary's complete 149-token list; version 2 intentionally expands it to 151. This also includes the bounded warning proposed in closed PR #16084.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.