Add first-class CODEOWNERS indexing and structural search

[Widthdom]

Problem

CODEOWNERS is high-value repository structure for review, security, and change-impact workflows, but cdidx does not index it.

This repository's .github/CODEOWNERS is reported as a language_support_candidate; files, search, symbols, and outline cannot inspect its patterns or owners.

Environment

• origin/main: c642bc4c3dda4ef659df8e6b699e10f61da3c836
• cdidx v1.44.3, locally built
• fresh repository index
• full net8.0 and net9.0 suites pass

Reproduction

dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll files \
  --db .cdidx/codeindex.db --path .github/CODEOWNERS --json

dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll search '@' \
  --db .cdidx/codeindex.db --path .github/CODEOWNERS --json

dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll languages \
  --extension CODEOWNERS --json

All return no indexed language/file/results, while status --json lists the file as an unknown extensionless language-support candidate.

Motivation

First-class CODEOWNERS data would let users and agents answer:

• Who owns a file or directory?
• Which patterns mention a team, user, or email?
• Which rules are broad, ownerless, or overridden later?
• Which reviewers are relevant to impact/dependency results?
• Does a sensitive directory lack an effective owner?

Proposed solution

Add an exact-filename codeowners structural language/extractor.

At minimum:

• recognize supported CODEOWNERS locations and exact filenames;
• index raw text for ordinary search;
• emit one ordered rule symbol per pattern;
• retain owners and source line as structured metadata;
• support files, search, outline, and symbols --lang codeowners;
• remove successfully indexed files from unknown-extension diagnostics.

Effective owner resolution for arbitrary paths can be included only if its precedence and syntax are implemented authoritatively; structural extraction must not pretend to implement authorization.

Focused PR scope

One PR should:

1. add exact-filename detection for CODEOWNERS;
2. implement a bounded line-oriented parser;
3. emit rule symbols with pattern, owners, path, order, and line;
4. preserve raw chunks and useful outline output;
5. add language capability/catalog/help documentation;
6. update unknown-extension classification;
7. expose minimal structured queries through existing symbol/field facilities;
8. add the bilingual changelog fragment.

Non-goals

• Do not call GitHub to resolve users or teams.
• Do not treat CODEOWNERS as an authorization boundary or infer permissions.
• Do not add every hosting provider's ownership format in this PR.
• Do not advertise reference/call graph support without meaningful semantics.

Implementation guidance and cautions

• GitHub searches .github/, repository root, then docs/; document whether cdidx indexes all candidates and which one is effective.
• Preserve rule order because the last matching rule controls effective ownership.
• Handle comments, escaped leading #, whitespace, ownerless rules, multiple owners, CRLF, and malformed lines conservatively.
• CODEOWNERS differs from full .gitignore; do not add unsupported negation or character-range semantics.
• Bound patterns and owner strings with existing symbol/string budgets.
• Use a natural rule symbol kind unless a new kind is justified.
• Keep @org/team, @user, and email owners searchable without treating them as code references.
• If path matching is added, make precedence and unsupported syntax explicit and tested.

Required tests

• .github/CODEOWNERS, root CODEOWNERS, and docs/CODEOWNERS.
• Multiple candidate locations and effective-file precedence.
• Comments, blank lines, escaped markers, whitespace, and CRLF.
• Multiple/team/email owners and ownerless rules.
• Duplicate and overlapping rules where later lines win.
• Invalid/oversized lines with bounded diagnostics.
• Files, raw search, symbols, and outline.
• Status unknown-extension behavior.
• Incremental add/update/delete and branch-switch indexing.

Acceptance criteria

• .github/CODEOWNERS appears in files and raw search.
• Every valid ownership rule has a stable ordered structural result.
• Pattern and owners are directly queryable.
• Rule order and effective-file semantics are documented.
• CODEOWNERS is no longer reported as unknown after a current full scan.
• No graph capability is advertised unless intentionally implemented.

Duplicate preflight

No open or closed issue was found for first-class CODEOWNERS indexing or structural search.