MAP keys serialize intederministically

[agavra]

Different implementations of Java's Map serialize differently. This means that we could have the same map serialize to two different bytes representations, and therefore get sent to different partitions. This was discovered with the following pull query test:

Note: this test will fail in master because we have not yet implemented non-literal keys. It was run on a test branch where I am implementing that feature

    {
      "name": "non-windowed - map keys",
      "properties": {
        "ksql.key.format.enabled": true
      },
      "statements": [
        "CREATE STREAM INPUT (ID MAP<STRING, STRING> KEY, IGNORED INT) WITH (kafka_topic='test_topic', format='JSON');",
        "CREATE TABLE AGGREGATE AS SELECT ID, COUNT(1) AS COUNT FROM INPUT GROUP BY ID;",
        "SELECT ID, COUNT, ROWTIME FROM AGGREGATE WHERE ID=MAP('F2':='0', 'F1':='1');"
      ],
      "inputs": [
        {"topic": "test_topic", "timestamp": 12345, "key": {"F1": "1", "F2": "1"}, "value": {}},
        {"topic": "test_topic", "timestamp": 12365, "key": {"F1": "1", "F2": "0"}, "value": {}}
      ],
      "responses": [
        {"admin": {"@type": "currentStatus"}},
        {"admin": {"@type": "currentStatus"}},
        {"query": [
          {"header":{"schema":"`ID` MAP<STRING, STRING> KEY, `COUNT` BIGINT, `ROWTIME` BIGINT"}},
          {"row":{"columns":[{"F1": "1", "F2": "0"}, 1, 12365]}}
        ]}
      ]
    },

The lookup could not succeed because the serialized key did not match the serialized where clause value.