Post-1.3: tray actions should reflect service-down state

Manual 1.3 testing: `capsem status` showed the service stopped, but the tray still offered actions like dashboard/session listing that cannot work while the daemon is down.

Expected:
- When service is down, tray should clearly show service unavailable.
- Actions that require the daemon should be disabled or replaced with Start/Retry.
- No dashboard/session action should lead to confusing API errors when the service is known down.

Post-1.3 polish unless it blocks service start.

Start by tracing the tray action handling and how it consumes the service state reported by `capsem status`. Verify behavior with the daemon stopped: the tray should show that the service is unavailable, prevent dashboard and session actions from producing API errors, and offer Start or Retry instead.

Written by the indexing model from the issue text.