On-demand ServerGroup cache refresh causes temporary instance data loss

When on-demand cache refresh is triggered via /cache/aws/ServerGroup endpoint in Clouddriver, instance data becomes temporarily invalid for 5-10 seconds, typically occurring 5-10 seconds after the refresh completes. The instances key returns empty or null values instead of actual instance IDs. This timing correlates with the Redis caching agent scheduler interval (redis.poll.intervalSeconds).

Apply fix from clouddriver PR #5033 which adds authoritativeTypes to ClusterCachingAgent on-demand results to prevent putCacheResult from replacing instances cache. As a workaround, add retry logic in Orca tasks to check if knownInstanceIds contains null and return TaskResult.RUNNING. Increase redis.poll.intervalSeconds to 60 to extend the timing window for diagnosis.