[codex] Add desktop hands-free voice mode

[aj47]

Summary

• Adds shared hands-free voice config, phrase matching, and controller state utilities in @dotagents/shared so mobile and desktop use the same wake/sleep behavior.
• Wires desktop settings, config normalization, and remote settings API fields for hands-free voice.
• Adds a non-modal desktop hands-free voice dock in the main app layout so the existing app UI stays navigable while voice remains armed.
• Routes hands-free speech to the currently focused session/agent when available, falling back to a new session when nothing is focused.
• Keeps the original one-shot desktop voice recording flow available when hands-free mode is off.

Why

The first desktop implementation started correctly, but the blocking modal hid the active agent UI after execution began. Hands-free voice now lives as an app-level control surface instead of taking over the screen, so the user can continue watching agent progress, navigating sessions, and talking to whichever agent/session is focused.

Validation

• pnpm --filter @dotagents/shared test
• pnpm --filter @dotagents/shared typecheck
• pnpm --filter @dotagents/core typecheck
• pnpm --filter @dotagents/desktop typecheck
• pnpm --filter @dotagents/desktop exec vitest run src/main/remote-server.routes.test.ts src/renderer/src/pages/settings-general.langfuse-draft.test.tsx src/renderer/src/pages/sessions.in-app-actions.test.ts src/renderer/src/components/app-layout.session-retention.test.ts src/renderer/src/components/session-action-dialog.voice.test.tsx
• pnpm --filter @dotagents/mobile test:vitest
• pnpm build:shared
• git diff --check

Notes

Unrelated untracked local files were intentionally left out of this PR: the desktop PDF files and root tsconfig.json.