tests: broaden fixture corpus quality coverage

[gaelic-ghost]

Summary

• add synthetic near-miss and longer-body records to the Gutenberg fixture corpus
• cover focused-vs-scattered all-term ranking and longer-body snippet selection in in-memory FetchKit tests
• add SearchKit parity coverage for the same fixture behavior
• reward tighter all-term evidence in the in-memory ranker so focused passages beat scattered matches
• document why the synthetic fixture records exist

Verification

• swift test --filter FixtureCorpusQualityTests
• swift test --filter SearchKitFetchIndexTests/testSearchKitFetchIndexMatchesFixtureCorpusNearMissAndLongBodyBehavior
• swift test
• scripts/repo-maintenance/validate-all.sh

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0305735ba0

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Avoid quadratic index-distance scans in term location loop

The new compactness scorer makes every .allTerms query walk each matching document body via termLocations, and this loop computes distance(from: startIndex, to:) for every hit. On long texts with frequent terms (for example, a common word appearing thousands of times), those repeated distance calculations accumulate to roughly O(n²) work per term, which can make in-memory search latency spike substantially compared with the previous constant-time scoring path. Converting the search text to a random-access representation once (or tracking offsets incrementally) avoids this regression.

Useful? React with 👍 / 👎.