Cross-file duplicate Quote rows caused by pre-correction StableId hashing #220

Title

Cross-file duplicate Quote rows caused by pre-correction StableId hashing

Body

While doing a full manual title-consistency review across the bundled quote sources (#181/#217), found an architectural gap that #181's rule/alias mechanisms cannot fix: QuoteIdentity.StableId hashes normalize(quote text) + "|" + normalize(source text) using the raw, pre-correction source text. When the same real-world quote appears in two different bundled files under two different raw source-text spellings, this produces two different quote ids for what is genuinely the same quote — and no per-quote or per-source correction can merge them after the fact, because the id itself was already computed from the "wrong" text before any correction ever runs.

Two confirmed live instances:

1. The Godfather Part II: NikhilNamal17 quote 3eb18c68-6211-ed48-b54a-6ebbb42d6c79 ("Keep your friends close, but your enemies closer.", raw source "The Godfather II") and vilaboim quote 9ee90d54-cf0e-014c-b4a1-4378ddd2d791 (identical quote text, raw source "The Godfather Part II"). Minimal per-source conflict-resolution rule file + curated field-override preload #181 added a SourceAliasRule so NikhilNamal17's copy now resolves to the correct "The Godfather Part II" Source — but the two rows remain two separate Quote entries, both linked to the same (now-correct) Source, because their ids were already fixed at conversion time from the differing raw source text.
2. The Shawshank Redemption: NikhilNamal17 quote 7e53658c-0c3a-6546-8c12-c5e4af23c9f8 (raw source "Shawshank Redemption") and another NikhilNamal17 quote e97a8197-657f-ce43-8032-4edd8e3549a3 (raw source "The Shawshank Redemption") have byte-for-byte identical quote text ("Hope is a good thing, maybe the best of things, and no good thing ever dies.") but different ids, for the same reason.

Why this is a different problem from #181's own scope: #181's SourceAliasLookup and ConflictResolutionRule both operate on an already-assigned entity id — they correct which Source a quote links to, or which field value a quote displays, but never merge two genuinely-duplicate Quote rows that already have different ids.

Possible directions for a future issue (none decided, needs its own design):

• Canonicalize source text (via the same alias mechanism, or a shared normalization step) before QuoteIdentity.StableId is computed, so an id-generation pass run after aliases exist would naturally produce matching ids. This likely only helps for future/regenerated data, not already-shipped ids.
• A manual Quote-level merge mechanism — an explicit "these two ids are the same quote, keep one, soft-delete the other" operation, usable after the fact regardless of when the ids were assigned.

Not designed or scoped yet. See docs/milestones/data-import-sources/181-minimal-conflict-resolution-rule-file-plan.md for the full context of the review this came out of.