Kafka 2.4.0

[solsson]

Highlights for us include:

• Consume from nearest replica, https://issues.apache.org/jira/browse/KAFKA-8443. Until now Multizone rack awareness #41 was good for resilience, but now it might also help performance.
• We could have a leader election job, https://issues.apache.org/jira/browse/KAFKA-8286
• Dynamic log levels, could also be a job, https://issues.apache.org/jira/browse/KAFKA-7800
• We could possibly speed up some automated group consumer setups thanks to https://issues.apache.org/jira/browse/KAFKA-7471
• Topic creation jobs no longer need to decide on replication factor and partitions, https://issues.apache.org/jira/browse/KAFKA-8305, which helps when we want to reuse automation across prod and development

[solsson]

I'll scope out both client.rack for test producers/consumers and 2.4.0 admin CLI from this PR because the former depends on reading node labels (like kafka's init-config or https://github.com/Yolean/kubernetes-kafka/compare/pod-ip-listeners or https://docs.google.com/document/d/1fH64mEfZH597luup-ZBfBNkiTVjfoFYGEa-G1G_TM6A/edit#, https://github.com/bjhaid/node-labels-to-files etc) and the latter depends on laborious native-compile work (or going back to full JVMs for admin commands, i.e. slow and expensive).