Propagate unknown-origin coverage into ad hoc search count authority

[Widthdom]

Problem

Ad hoc search --count reports authoritative_count: true and exit 0 after origin filtering has excluded matches whose C# lexical context is unknown. The count describes accepted indexed search rows, but cannot establish the complete code-origin count or authoritative absence.

This was observed while validating #5348. That issue adds bounded lexical continuation; propagating classification coverage into the existing ad hoc count-output path is a separate follow-up. Regex find already retains its partial/unknown outcome, and recipe materialization already observes origin coverage before filtering.

Reproduction

Using the repository-built Debug/net8.0 cdidx 1.49.0 and an index of 1ad7e604d336c38b7df1508e9ac339c34b9b230f:

dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll search return --path src/CodeIndex/Cli/QueryCommandRunner.SearchRecipes.cs --origin code --format count --json --db .cdidx/codeindex.db
dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll search return --path src/CodeIndex/Cli/QueryCommandRunner.SearchRecipes.cs --origin unknown --format count --json --db .cdidx/codeindex.db
dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll find return --regex --path src/CodeIndex/Cli/QueryCommandRunner.SearchRecipes.cs --origin code --format count --json --db .cdidx/codeindex.db

The first command reports count 58, degraded: false, authoritative_count: true, and exit 0, although matches after the first lexical window remain unknown. The explicit unknown query confirms such search rows exist. The regex-find control reports 111 unknown-origin occurrences, origin_classification_complete: false, authoritative_count: false, and exit 11. Search counts indexed result rows whereas find counts occurrences; these numeric counts are not interchangeable.

The one-pass classification and ad hoc count-output behavior are unchanged by the #5348 implementation. The observation was made on macOS arm64 while testing that implementation, using the baseline index above.

Expected behavior and validation

• Observe unknown candidate origins before applying include/exclude filters and before output selection.
• Report incomplete/non-authoritative counts and actionable classification recovery guidance, with the shared partial-exit opt-in semantics.
• Preserve authoritative complete zero results and distinguish intentional output selection from incomplete classification.
• Check ordinary count, count format, grouped/count-by and named-query routes for the same propagation gap; keep recipe and regex-find controls.
• Add a small C# fixture with executable matches before and after a lexical window, plus malformed/missing-context controls; compare one-pass and explicitly continued classification on net8/net9.

Duplicate checks: open issues searched for search origin authoritative and "origin" "count"; no issue specifically tracking ad hoc count authority was found.