codewhale doctor: detect orphaned legacy .deepseek state that hasn't been migrated

[Hmbown]

Problem

codewhale doctor doesn't report whether legacy ~/.deepseek/ state directories exist but haven't been migrated to ~/.codewhale/. Users who upgraded and see empty sessions have no built-in way to diagnose that their data is still sitting in the old location.

Desired behavior

codewhale doctor should check each state subdirectory and report:

• If ~/.deepseek/<subdir> exists AND ~/.codewhale/<subdir> does NOT exist → warn that legacy data is unmigrated
• If both exist → note that some data may still be in the legacy location
• If only ~/.codewhale/<subdir> exists → all clear

Subdirectories to check: sessions, tasks, skills, slop_ledger, trophies, catalog, config.toml, settings.toml, mcp.json.

Example output:

⚠ Legacy state found but not migrated:
  ~/.deepseek/sessions/ → not yet in ~/.codewhale/sessions/
  Run CodeWhale once to auto-migrate, or manually:
    cp -r ~/.deepseek/sessions/* ~/.codewhale/sessions/

Context

• The migration logic exists in crates/config/src/lib.rs (migrate_legacy_state_dir(), resolve_state_dir(), ensure_state_dir())
• doctor already reports active state root and legacy root status (from P0 v0.8.44: migrate app state to ~/.codewhale with ~/.deepseek compatibility bridge #2011), but doesn't compare per-subdirectory
• This would also serve as a self-healing hint: running codewhale once triggers ensure_state_dir() which calls migrate_legacy_state_dir()

Related

• Sessions appear lost after upgrade: .deepseek → .codewhale read-path doesn't fall back to legacy sessions #3724 — session read-path fallback
• Provider picker: "needs-auth / auth:missing" is inaccurate for API-key providers (and reads as a readiness regression) #3725 — user-visible migration notice
• P0 v0.8.44: migrate app state to ~/.codewhale with ~/.deepseek compatibility bridge #2011 — original migration plan (v0.8.44)