[Bug] The combination of Command and Aggregate operators causes colmun verification to fail

### Code of Conduct

- [X] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct)

### Search before asking

- [X] I have searched in the [issues](https://github.com/apache/kyuubi/issues?q=is%3Aissue) and found no similar issues.

### Describe the bug

test code:
```
test("auth: Command and Aggregate operators causes colmun verification to fail") {
val db = "default"
val table = "src"
val col = "key"
val create = s"CREATE TABLE IF NOT EXISTS $db.$table ($col int, value int) USING $format"
val select = s"CACHE TABLE TEST AS SELECT key FROM $db.$table group by key"
withCleanTmpResources(Seq((s"$db.$table", "table"))) {
doAs("bob", sql(create))
doAs("kent", sql(select).collect())
}
}

```


explain :
```
CacheTableAsSelect TEST, SELECT key FROM default.src group by key, false, true
+- Aggregate [key#17], [key#17]
+- SubqueryAlias spark_catalog.default.src
+- HiveTableRelation [`spark_catalog`.`default`.`src`, org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe, Data Cols: [key#17, value#18], Partition Cols: []]
```

This paln did not generate **project** operator, causing PrivilegesBuilder's mergeProjection to merge all columns of HiveTableRelation.

### Affects Version(s)

master

### Kyuubi Server Log Output

_No response_

### Kyuubi Engine Log Output

_No response_

### Kyuubi Server Configurations

_No response_

### Kyuubi Engine Configurations

_No response_

### Additional context

_No response_

### Are you willing to submit PR?

- [X] Yes. I would be willing to submit a PR with guidance from the Kyuubi community to fix.
- [ ] No. I cannot submit a PR at this time.

Start by running the test case from the issue and tracing PrivilegesBuilder.mergeProjection for the shown CacheTableAsSelect and Aggregate plan. Check how the HiveTableRelation columns are handled during verification; done when the reproduction no longer merges unselected columns and the test passes.

Written by the indexing model from the issue text.