test: document rule-extraction case, punctuation and Unicode behavior (#2)

[Cid-oe]

Summary

Resolves #2.

Documents and asserts RuleExtractor's regression behavior across whitespace trimming, sentence-final punctuation, repeated punctuation, English case normalization, non-ASCII/Unicode letters, mixed CJK/Latin phrasing, and unsupported edge cases.

Changes

• Regression Tests (src/extract.rs):
  • Added table-driven test rule_extractor_behavior_matrix asserting positive extractions for:
    • Leading and trailing whitespace handling.
    • Sentence-final punctuation (., !, ?).
    • Repeated punctuation (!!!???).
    • Full-width Chinese punctuation and lead word stripping (喝).
    • Mixed English case normalization (ALICE PREFERS VIM -> alice, prefers, vim).
    • Capitalized multi-word predicates (Bob Works At Microsoft.).
    • Non-ASCII German umlauts (Jürgen likes München) and accented characters (Renée lives in Paris).
    • CJK mixed with Latin names (Bob 住在 上海).
  • Added explicit negative test cases asserting empty extraction on:
    • Unsupported/unmodeled verbs.
    • Passive constructions (Rust is liked by Alice).
    • Self-referential statements (Alice likes Alice).
    • Trailing stop words reducing object to empty.
• Documentation (docs/guide.md):
  • Added a Markdown table summarizing the supported and unsupported behavior matrix.
  • Clarified that rule extraction is intentionally limited pattern matching rather than general NER or semantic understanding.
• Chinese README (README.zh-CN.md):
  • Linked directly to docs/guide.md#rule-extraction-behavior-matrix.

Validation

• cargo test passes (61/61 library tests, all CLI & integration tests green).
• cargo clippy --all-targets --all-features -- -D warnings passes with zero warnings.
• git diff --check passes cleanly.