Unify owned-process lifecycle state

@bzawisto is already working on this.

Since Sep 11, 2026.

## Outcome

Private ERs and verifiers share child-process ownership and teardown bookkeeping.

## Scope

[Private ERs](https://github.com/magicblock-labs/redsuite/blob/06c80157f8bc87c39ac48fd59367a8e9280d2db9/crates/redsuite-core/src/topology/private.rs#L118) and [verifiers](https://github.com/magicblock-labs/redsuite/blob/06c80157f8bc87c39ac48fd59367a8e9280d2db9/crates/redsuite-core/src/topology/replicated.rs#L160) duplicate child/PID state, termination, resource-record updates, and drop handling. Their already-stopped finish behavior differs.

Introduce a small internal owned-process component. Keep readiness, identity leases, verifier reconnection, and storage-reset policy in their respective topology types. This follows the completed module decomposition in #36; preserve Linux behavior and coordinate macOS process support with #126.

## Acceptance criteria

- [ ] Process ownership and resource-record transitions have one implementation.
- [ ] Explicit finish handles running and already-stopped processes consistently, including intended storage cleanup.
- [ ] Relaunch and teardown failures remain visible in resource auditing.
- [ ] Graceful shutdown, hard-kill, restart timing, and topology-specific reset semantics remain intact.
- [ ] Keep the ownership refactor narrow; do not introduce a broad topology redesign or preclude the macOS support tracked in #126.