Deduplicate ProcessSupervisor liveness-readiness success logs

Context

After an orchestrator restart, the parent supervisor repeatedly emitted the same INFO line roughly every cooldown interval:

[ProcessSupervisor] lms server (LM Studio CLI) readiness hook completed successfully after liveness confirmation.

Child daemons may stay verbose, but the parent ProcessSupervisor log is the operator-facing signal stream. Repeating a stable success confirmation every poll makes real warnings and failures harder to spot.

Problem

ProcessSupervisorService.runLivenessReadinessHook() logs the successful liveness-confirmed readiness hook every time the fire-and-exit liveness path observes an already-up service. For stable readiness, that is state-level noise, not a new event.

Proposed Shape

• Keep the first successful readiness transition visible.
• Suppress duplicate INFO success lines while the same task remains healthy.
• Reset the suppression after degraded or failed readiness so a later recovery success is visible.
• Preserve WARN/ERROR visibility for degraded/failure paths.
• Keep the fix in ProcessSupervisorService; the orchestrator remains a thin scheduler.

Acceptance Criteria

• Repeated successful runLivenessReadinessHook() calls for the same healthy task do not flood parent supervisor logs.
• The first successful liveness-confirmed readiness hook still logs once.
• A degraded or failed readiness result resets the success guard so the next successful recovery logs once.
• Task health outcomes are still recorded for every readiness check.
• Unit coverage exercises the liveness-confirmed readiness hook success/degraded/recovery behavior.