We should design and prototype support for node pools. Node pools should be different groups of Kafka brokers with different configurations. That should allow to be scaled independently and be able to move brokers between them.

They should give us more flexibility for configuring the clusters, but also to support KRaft clusters with different architectures.