feat: add board defaults for model, reasoning effort and preset

[Fnckerpoi]

Summary

Add board-level defaults for the model, reasoning effort, and Agent preset used by newly created tasks.

• Add model/effort/preset selectors to Board Settings using the existing runtime catalogs and theme styles, with Chinese and English copy.
• Persist and validate defaultModel (including reasoningEffort) and defaultPresetId with the ledger; preserve them across restarts and JSON import/export.
• Materialize defaults onto new tasks through both the GUI create route and taskboard_create, without rewriting existing tasks when settings change.
• Keep explicit task/template selections ahead of board defaults. The GUI can explicitly follow deployment defaults using null create payloads, rather than accidentally inheriting the board default again.
• Reset reasoning effort when switching models and preserve saved selections when catalog entries are unavailable.
• Keep remembered last-model behavior when no board model default is configured.
• Include updated documentation and rebuilt, tracked Host/Client artifacts.

Verification

• npm run typecheck
• npm test -- --maxWorkers=2 — 28 test files, 378 tests passed
• npm run build
• git diff --check

Seven new tests cover settings UI save/reset behavior, unavailable catalogs, template/default precedence, explicit deployment-default overrides, Host route and tool creation, settings validation, and import behavior. The existing persistence test also now covers model/effort/preset defaults.

The first full run at default worker concurrency hit three scheduler lifecycle timing failures. Those tests passed when run independently (both the working tree and untouched HEAD baseline), and the complete suite passed with two workers. No scheduler code was changed.

Live Desktop plugin reload and real-browser visual verification were not performed; UI interactions were verified in jsdom.