Reduce container image sizes by consolidating operations per layer

See https://github.com/debezium/container-images/pull/312

Per Gunnar, he noticed that we unnecessarily increase the size of the Kafka image by doing ops on the/kafka directory in separate RUN steps. By merging them together into a single step, we save about 2 * 100 MB (first row is with the change, second row is the current upstream 2.0 image):

REPOSITORY                       TAG               IMAGE ID       CREATED          SIZE
quay.io/debezium/kafka           2.0               039ddb26fd88   20 minutes ago   665MB
<none>                           <none>            985dc0be200e   58 minutes ago   876MB