fix(processor): delimit sentinel escape fallback

[yisding]

Motivation

• Prevent corruption of user input when multi-character private-use escape tokens are chosen and a restore operation can match across an original private-use character / escaped-token boundary.

Description

• Wrap multi-character private-use escape token bodies in an absent noncharacter delimiter so restore matches are always aligned to whole escape tokens and cannot begin inside adjacent original private-use characters.
• Add _NONCHARACTER_DELIMITER_RANGES, _iter_delimited_private_use_tokens, _iter_noncharacter_delimiters, and _absent_noncharacter_delimiter helpers, and update _build_sentinel_escape_tables to prefer single-codepoint escapes and fall back to delimited multi-codepoint tokens when necessary.
• Make the compiled restore_re alternation length-sorted for safe left-to-right atomic restoration and keep str.translate/regex-based round-trip semantics.
• Add a regression test test_sentinel_restore_does_not_match_across_original_private_use_boundary to tests/regression/test_library_review_fixes.py that exercises the cross-boundary case via both Processor.process() and the public segment_clean() path.

Testing

• Ran the targeted regression file: uv run pytest tests/regression/test_library_review_fixes.py -q and it passed (59 passed).
• Ran lint/format checks: uv run ruff check . and uv run ruff format --check . and both passed.
• Ran full test suite with coverage: uv run pytest --cov=sentencesplit tests/ --color yes and the suite passed (1717 passed, 8 xfailed).

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 883c5a6c51

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".