[cli-status-doctor-semantics] 1-3-2 corrupted diagnostic bundle v2

[jununfly]

Context

Roadmap: cli-status-doctor-semantics
Node: 1-3-2 corrupted诊断bundle后续版本

This was discovered while unifying CLI/status/doctor graph-health semantics.
The current shipped slice makes zcodegraph status classify an unreadable
existing database as corrupted and print recovery commands without
auto-cleaning or auto-rebuilding.

Problem

zcodegraph doctor --bundle currently depends on status/DB paths that may try
to open .zcodegraph/zcodegraph.db. For a corrupted SQLite database, that can
fail before a useful diagnostic bundle is created.

Desired outcome

Add a doctor bundle path for corrupted graph databases that does not require
opening the malformed DB. It should collect replay-safe metadata such as:

• DB file presence, size, mtime, and sanitized open error
• diagnostic record availability (last-run, last-failure)
• project/root hash and privacy-preserving environment metadata
• exact recovery commands

It must not upload anything automatically and must not include source code or
plaintext source paths by default.

Acceptance gate

• A corrupted .zcodegraph/zcodegraph.db can produce a local diagnostic bundle.
• The bundle path does not call CodeGraph.open on the corrupted DB.
• Tests cover the malformed DB path and privacy policy.
• zcodegraph status remains non-destructive and continues to print recovery
  commands rather than auto-cleaning.

[jununfly]

Implemented by PR #674 and merged into main.\n\nThis closes the corrupted diagnostic bundle v2 slice: zcodegraph doctor --engine rust-hybrid --bundle can now create a local, privacy-preserving bundle when the graph database is malformed, without opening the broken DB or auto-cleaning it.\n\nVerification recorded in the PR: npx vitest run __tests__/status-json.test.ts __tests__/rust-hybrid-doctor.test.ts, npm run build, and git diff --check.