Stream GROUP BY on ROWKEY does unnecessary repartition step

[big-andy-coates]

The following QTT test would fail due to the presence of a repartition topic. However, the GROUP BY is on ROWKEY and therefore no re-partition is required.

{
      "name": "ROWKEY (stream->table) - without repartition",
      "comment": [
        "Clone of test 'ROWKEY (stream->table)' but checking no repartition topic is created post v5.4"
      ],
      "statements": [
        "CREATE STREAM TEST (ignored VARCHAR) WITH (kafka_topic='test_topic', value_format='DELIMITED');",
        "CREATE TABLE OUTPUT AS SELECT ROWKEY, COUNT(*) FROM TEST GROUP BY ROWKEY;"
      ],
      "inputs": [
        {"topic": "test_topic", "key": 1, "value": "-"},
        {"topic": "test_topic", "key": 2, "value": "-"},
        {"topic": "test_topic", "key": 1, "value": "-"},
        {"topic": "test_topic", "key": 2, "value": "-"},
        {"topic": "test_topic", "key": 1, "value": "-"}
      ],
      "outputs": [
        {"topic": "OUTPUT", "key": "1", "value": "1"},
        {"topic": "OUTPUT", "key": "2", "value": "1"},
        {"topic": "OUTPUT", "key": "1", "value": "2"},
        {"topic": "OUTPUT", "key": "2", "value": "2"},
        {"topic": "OUTPUT", "key": "1", "value": "3"}
      ],
      "post": {
        "topics": {
          "blacklist": ".*-repartition"
        }
      }

[big-andy-coates]

Related to #3366

[big-andy-coates]

I've now fixed this.