fix: harden container shutdown cleanup

[cristibleotiu]

What changed: abort container restarts when runtime cleanup fails; preserve process, thread, and fixed-volume handles for retry; stop Docker log readers after container stop to avoid false cleanup failures.

Why: prevent stuck or duplicate runtimes during restart and shutdown paths.

PR summary:

• Made container shutdown failures retryable instead of getting stuck forever after one failed cleanup.
• Made manual STOP safer: it only persists “paused” after cleanup actually succeeds.
• Added explicit RESTART behavior that cancels a pending failed manual stop.
• Prevented config changes from accidentally overriding a pending manual stop.
• Hardened Docker cleanup so failed stop/remove keeps the container handle for retry.
• Hardened fixed-volume cleanup so mounted loop devices are not leaked when metadata is missing, malformed, or stale.
• Added compatibility fallback code so edge can still clean tunnel subprocess trees even if deployed before the
  matching core change.
• Increased the plugin stop timeout for container apps because Docker, tunnels, log readers, and fixed volumes can
  legitimately need longer cleanup time.
• Resolved the develop conflict by preserving the newer volume-sync lifecycle behavior and integrating it with the
  shutdown hardening.