SwimlaneDispatcher configurable - hashing algorithm, threadpool size

[cer]

Currently,

• the SwimlaneDispatcher parallelizes message processing by partition.
• Messaged from different topics with the same partition are processed by the same thread.
• As a result, the degree of parallelism is limited

Proposal:

• Make the (topic, partition) -> thread algorithm configurable

[cer]

To configure the SwimlaneBasedDispatcher, configure an @Bean that implements TopicPartitionToSwimLaneMapping.

For example, see SwimlanePerTopicPartition

[cer]

One more thing to consider: pass the message key to toSwimLane(TopicPartition topicPartition) to enable parallelism within a TopicPartition.

[cer]

I believe that MultipleSwimlanesPerTopicPartitionMapping is unnecessarily complicated. A simpler implementation could simply hash(topic and messageKey) % desiredNumberOfSwimlanes. This will preserve ordering of all messages published to a topic (e.g. Aggregate type) with the same message key (e.g. aggregate ID)