fix: keep shared Chrome alive across agent exits - #16

What

• detach managed Chrome from its launching Windows controller so PID lease ownership can pass between concurrent agents
• retry transient EPERM and EBUSY recovery-file replacements automatically
• make the documented normal and concurrent workflow zero-config; ASK_PRO_AGENT_ID remains only for explicit profile isolation

Why

The shared browser lease correctly left Chrome running for a peer, but chrome-launcher spawned Chrome as a non-detached Windows child. When the first controller exited, its host process tree could still terminate Chrome under the second agent. Transient Windows rename failures could then prevent saved browser metadata from making that interrupted answer recoverable.

Impact

Agents still run ask-pro, attach context, and submit the prompt. There are no new flags, environment variables, settings, profiles, or services.

Checks

• plugin validator
• pnpm run build
• pnpm run lint
• pnpm test (287 passed)
• pnpm run format:check
• pnpm pack --dry-run
• Review Suite fast + exact-head simplification closure: rvw_002abc85 (clean)