Regeneration of DISABLE operations when setting instance operations

### Describe the bug
There is a bug in the instance operation handling logic that causes an infinite cycle of regenerated DISABLE operations when applying operations like EVACUATE to an instance with HELIX_ENABLED=false in legacy fields. This prevents proper instance lifecycle transitions and causes unexpected behavior in instance swapping scenarios.

### To Reproduce
1. Create an instance in the cluster.
2. Set DISABLE operation with USER source:
Apply to InstanceConf...
`curl -X POST "http://:/admin/v2/clusters//instances/?command=setInstanceOperation&instanceOperationSource=USER&instanceOperation=DISABLE"`
3. Set EVACUATE operation with AUTOMATION source:
Apply to InstanceConf...
`curl -X POST "http://:/admin/v2/clusters//instances/?command=setInstanceOperation&instanceOperationSource=AUTOMATION&instanceOperation=EVACUATE"`
4. Set EVACUATE operation again with AUTOMATION source
5. Check the instance configuration in ZooKeeper

After the second EVACUATE operation, a new DISABLE operation is generated with:
1. Updated timestamp (more recent than the EVACUATE timestamp)
2. Added LEGACY_DISABLED_TYPE field
3. The cycle repeats with each subsequent evacuate operation

### Expected behavior
The DISABLE operation should not be regenerated with new timestamps when setting a different operation. Instance operations should be updated consistently without creating an infinite cycle.

### Additional context
RCA:
The regeneration happens because:
1. getActiveInstanceOperation() returns the last operation (EVACUATE)
2. EVACUATE is in INSTANCE_DISABLED_OVERRIDABLE_OPERATIONS
3. This causes getInstanceOperation() to override it with a newly generated DISABLE operation

Start with getActiveInstanceOperation() and getInstanceOperation(), then reproduce the two EVACUATE requests through the setInstanceOperation API and inspect the instance configuration in ZooKeeper. Trace why EVACUATE causes a new DISABLE operation when HELIX_ENABLED=false in legacy fields. Done means repeated EVACUATE operations no longer regenerate DISABLE timestamps or LEGACY_DISABLED_TYPE entries, and instance transitions proceed normally.

Written by the indexing model from the issue text.