feat(library): add curated scopes and trust metadata

[YiShu5]

Summary

• split the public deep-content library into manually curated AI and cross-domain scopes
• replace precise public scores with conservative AI rating labels
• expose explicit source and processing metadata without inferring verification
• unify runtime and strict archive validation with fail-closed behavior

Product behavior

• /library defaults to AI 主内容 and supports 跨界深读
• scope membership is maintained by stable metadata IDs
• unmapped records are not silently assigned to either scope
• public pages display AI 初评 · <等级> instead of exact content scores
• search similarity remains a separate N% 相似 signal
• detail pages show only explicitly recorded source and processing fields
• the empty state is now 暂无已完成整理的内容

Safety and consistency

• runtime and strict validation share the same archive directory, metadata ID, public-status, and duplicate-ID rules
• only _duplicates_quarantine is excluded by exact directory name
• invalid metadata is diagnosed and omitted without taking down public pages
• duplicate public archive IDs are excluded as a complete conflict group
• no archive data or topic identity mappings were modified
• search recall, ranking, embeddings, and 问降噪 behavior were not changed

Validation

• blog tests: 375 / 375
• deployment tests: 29 / 29
• total Python tests: 404 / 404
• Node checks: 5 / 5
• Python compileall: passed
• cached diff check: passed
• scope mapping: 45 records, 32 AI, 13 cross-domain
• production archive read-only validation: 45 public completed records, no unmapped, stale, duplicate, or invalid records

Files

• 17 files
• 2,345 insertions
• 98 deletions

Non-blocking note

When every archive record is rejected by fail-closed validation, the existing empty-list cache behavior may cause later requests to rescan and repeat diagnostic logs. This does not affect correctness or the current valid 45-record production inventory.

Out of scope

• feedback collection
• accounts, subscriptions, bookmarks, or private knowledge bases
• topic identity changes
• archive migration
• search algorithm changes
• deployment