fix: stop DestinationMigrationCoordinator cycling on unrelated cluster events

[thecodingshrimp]

Problem

DestinationMigrationCoordinator.clusterChanged() has no index-scope filter. Any cluster state event resets finishFlag and reschedules the migration job — even events completely unrelated to the scheduled-jobs index (e.g., datastream mapping updates from security-analytics). On a cluster where destinations have already been migrated, this causes an infinite Reset→Cancel→Perform cycle that generates continuous log noise.

Root cause

The clusterChanged callback resets finishFlag = false and reschedules unconditionally. Once migration completes and finishFlag = true, the next unrelated cluster event immediately restarts the cycle.

Fix

Add the same index-scope guard used by JobSweeper:
```kotlin
if (!event.indexRoutingTableChanged(ScheduledJob.SCHEDULED_JOBS_INDEX)) return
```
This ensures the coordinator only acts on routing-table changes to .opensearch-alerting-scheduled-jobs — the only index whose state is actually relevant to destination migration.

Testing

• Verified guard fires correctly: unrelated cluster events (mapping updates, etc.) return early without touching finishFlag or the migration scheduler
• Migration still completes correctly when .opensearch-alerting-scheduled-jobs routing changes