Improve adaptation quality evaluation

[Axelxrd]

Summary

• improve local adaptation heuristics for realistic long-form manuscripts, including dialogue speaker detection, specific location/prop extraction, and scene objective/conflict/turning point selection
• add a long-form original evaluation fixture and quality assertions for chapter coverage, dialogue density, characters, locations, props, scene function, and coverage scores
• add structured adaptation quality gates to schema/API/Web UI and render them in the quality overview

Validation

• python -m pytest -q --basetemp .pytest-tmp-quality-final2
• python -m ruff check .
• node --check src\novel2script\web\static\app.js
• python scripts\smoke_web_server.py
• git diff --check (CRLF warnings only)
• Edge + puppeteer-core: long-form conversion rendered 5/5 quality gates, all pass, with no horizontal overflow at 1440px or 390px