MessageConsumerKafkaImpl should allow for concurrent processing of messages from different topics (even if they have the same partition number)

[douggish]

Currently, it looks like MessageConsumerKafkaImpl passes a record's partition number as the "swimlane" to a SwimlaneBasedDispatcher:

eventuate-messaging-kafka/eventuate-messaging-kafka-consumer/src/main/java/io/eventuate/messaging/kafka/consumer/MessageConsumerKafkaImpl.java

Lines 42 to 44 in a8d2c99

	EventuateKafkaConsumerMessageHandler kcHandler = (record, callback) -> swimlaneBasedDispatcher.dispatch(new RawKafkaMessage(record.value()),
	record.partition(),
	message -> handle(message, callback, handler));

I believe this is to allow for concurrent processing of messages from different partitions while preserving ordering within a partition. This is desirable, and I think it works well if you are subscribed to a single topic.

However, it is possible to subscribe to multiple topics/channels, in which case it will dispatch messages with the same partition number to the same swimlane, even if they are from different topics. I don't think there are any ordering guarantees for messages in different topics (even if they have the same partition number), so MessageConsumerKafkaImpl should probably allow you to process messages from different topics concurrently. Otherwise messages that may be slow to process from (Topic A, Partition 1) will unnecessarily delay processing of messages from (Topic B, Partition 1).