Start a benchmark setup, then improve start times

[solsson]

The simplest setup I could think of, but staying with Java (where [kkv-example](https://github.com/Yolean/kafka-keyvalue/blob/master/kontrakt/kkv-example.yaml uses bash).

The idea is to run ./benchmark.sh which

• Starts an ephemeral kafka instannce
  • Yay, no zookeeper!
• Prepares a topic with lots of key-values
  • Parameters could be for example the number of messages and size of the key space
• Starts kkv from current source
• Starts an app that can receive "onupdate" hooks
  • There is such a start, but as a manual step echo'd by the script
• Measures time to readiness
• Maybe scoped up later to measures other latencies when producing to the topic
• Runs all of the above in a kubernetes cluster
  • There's a skaffold.yaml with the basics, but I got stuck on the choice of port-forward or in-cluster quarkus dev