refactor: Ctrl service-locator → app/Container DI (Phase 2, #234)

[pajoma]

Phase 2 of #234 — replace the Ctrl service-locator with a real composition root + dependency injection.

Depends on #235 (Phase 1) being merged.

Changes

• src/app/ new composition root:
  • container.ts — Container implements JournalController. Single-pass constructor (configSource, loggerFactory) builds the whole service graph in one go. Removes the two-phase initServices + ! non-null fields.
  • register.ts — command / code-action / cache-invalidation registration.
  • startup.ts — extension lifecycle (moved from vscode/startup.ts), .then() chain rewritten to async/await.
• Commands, UI providers and features now depend on the JournalController interface, never the concrete controller class.
• Expanded IConfiguration with the 11 methods consumers actually call (getNavigationMode, getScopeDefinitions, getWeeksPathPatternRaw, getWeeksFilePatternRaw, getWeeklySyncConfig, getTimeStringTemplate, getFileLinkInlineTemplate, getDailyLinkInlineTemplate, getBasePathForLocalOpen, getResolvedEntryPathForLocalOpen, getWeekPathPatternForLocalOpen).
• Moved WeeklySyncConfig type to model/. journal/paths.ts now types its config param as IConfiguration (drops a journal → vscode coupling).
• Deleted src/util/controller.ts and src/vscode/startup.ts.
• Tests construct new Container(cfg, () => logger) instead of new Ctrl(cfg); ctrl.initServices(logger).

Lifecycle decision (per #234 review)

Container is the single owner; all services are singletons (one instance per activation), matching the old single-Ctrl model. No separate state module.

Verification

• npm run compile-tests (tsc) ✔
• npm run compile (esbuild) ✔
• npm run lint ✔
• Full suite: 238 passing, exit 0

Acceptance (#234)

• No command/provider depends on concrete Ctrl/Container — all take JournalController.
• Single wiring point in app/container.ts.
• Two-phase init removed.

Refs #234.

🤖 Generated with Claude Code

[pajoma]

Brooks-Lint PR Review: Approved

Health Score: 92/100

Phase 2 of #234 is successfully implemented. This PR replaces the \Ctrl\ service-locator with a dedicated composition root in \�pp/container.ts, significantly improving the system's modularity and testability.

Key Findings:

• Inversion of Control: The service graph is now constructed in a single pass. Commands depend on the \JournalController\ interface, enabling true dependency injection.
• Robustness: The removal of two-phase init (\initServices) eliminates fragile non-null assertions across the controller.
• Configuration: While \IConfiguration\ remains large, this is a known transitional state to be addressed in Phase 3.
• Verification: 238 tests passed; zero regressions in behavior.

Recommendation: Merge immediately to unblock Phase 3 (Split Configuration).