Entity store: clean up orphaned index entries

Problem

When an entity is deleted from the etcd store but its index entry is not cleaned up, the reconciler logs a WARN + ERROR pair every 30 seconds, forever:

[WARN]  coordinator.etcdstore failed to get primary entity from etcd │ id: lsvd_mount-CXxPKn4o8A125BizS1AJe
[ERROR] coordinator.entityserver entity in index but not in store, skipping │ id: lsvd_mount-CXxPKn4o8A125BizS1AJe

On James's server this has been firing continuously since at least Feb 22 — over 40 days of noise at the time of discovery.

Expected Behavior

When the entity store encounters an index entry pointing to a nonexistent entity, it should:

1. Remove the stale index entry
2. Log a single WARN that it cleaned up an orphan
3. Not retry on subsequent reconciliation loops

Stretch

Add a startup validation sweep that checks all index entries against the store and cleans up any orphans found.