Add launchd-managed background shares

[icodesign]

Problem

An agent-launched remote-installer share could be killed when the temporary command session ended, leaving the install URL visible while the local origin and tunnel were already gone. The existing --timeout only controlled the worker after it had started; it did not give the process a lifetime outside the agent command session.

What changed

• Added macOS share --background, which registers a native worker with the user's launchd GUI domain and waits for the origin and tunnel to become ready before returning.
• Added finite-expiry enforcement for managed shares and structured ready output with a share ID.
• Added status, logs, and stop commands for managed share lifecycle control.
• Added managed session state, persisted stdout/stderr logs, launchd process-group ownership, and graceful SIGTERM handling.
• Updated the README, npm package documentation, and agent skill instructions to use the durable background workflow and to avoid temporary command sessions, &, or nohup for links that must survive an agent turn.

Validation

• cargo fmt -- --check
• cargo test --all-targets — 158 tests passed (119 library, 18 CLI, 3 Android, 6 availability, 3 download progress, 5 OTA, 4 range-request tests).
• cargo clippy --all-targets -- -D warnings
• Real launchd smoke tests: the launcher returned after registration while the native worker and install page remained available; normal expiry shut down the worker and tunnel; SIGKILL of the worker resulted in launchd process-group cleanup of the tunnel child and the session reported failure.