fix: repair optional integration suite for live extractor API

[Copilot]

Optional integration CI failed setup on every test: fixtures patched insight_extractor.tokenizer.AutoTokenizer (TYPE_CHECKING-only / lazy-loaded) and asserted phantom constructor flags plus obsolete ExtractResult fields.

Changes

• Integration fixtures — inject FakeModel / FakeTokenizer on _model / _tokenizer (no HF download, no AutoTokenizer patch)
• Test contracts — assert live API only: enable_dynamic_regex, regex_entities, dynamic_keyword_matches, semantic_keywords, key_sentences, KeywordStats
• CI job — rename off “full BERT pipeline”, drop unused HF cache, set HF_HUB_OFFLINE / TRANSFORMERS_OFFLINE
• Docs — README / SPEC / AGENTS / CLAUDE now describe offline orchestration, not stale mocks

extractor = InsightExtractor(
    seed_keywords=["ransomware", "CVE"],
    enable_dynamic_regex=True,
    output_dir=tmp_path,
)
extractor._model = FakeModel()
extractor._tokenizer = FakeTokenizer()
result = extractor.extract(sample_text)  # ExtractResult, not phantom fields

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-12T18:30:11.129567Z	71ac403	Draft marked ready

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.