Story 5.4b — Orchestrate cluster rebuild via forked solo-provisioner CLI child

**Epic:** #502 — Network Upgrade Workflow (Execute Phase)

(Split from #540, part 2 of 3.) Orchestrate the cluster teardown/reinstall by handing off to the
detached `solo-provisioner` CLI child process built in epic #500.

**Acceptance criteria:**
- After `PendingInfraUpgrade` is set, fork the detached CLI child (reusing #500's `Setsid`/`Setpgid`/`Release` mechanism) to run cluster teardown → reinstall → operator deploy → daemon restart.
- The CR phase remains `PendingInfraUpgrade` across the rebuild (etcd preserved on the `/opt/solo/weaver/sandbox/` hostPath, so the CR survives).
- The daemon process is stopped and restarted by the child; it must not assume continuity of in-memory state.
- Self-upgrade of the provisioner binaries, when required, is delegated to epic #500 (the CLI owns daemon-binary install).

**Blocked on:** epic #500 (self-upgrade detached-child mechanism — #523/#525/#526).
**Depends on:** #707 (5.4a infra detection).