Prevent crashes by ensuring topics are created

[mhratson]

addresses #390

Context

0. whenever a kafka-monitor starts monitoring a new cluster
1. it has to create topics (if enabled)
2. monitoring process then describes the topics and does its job

Problem

• due to possible delays between 1. and 2.
• step 2 may crash, as due to propagation delay, metadata may not be yet available
• this prevents kafka-monitor from starting as this may happen for multiple topics that may have to be created

Solution

1. after creation query topic metadata
2. use retries and backoff to allow enough time for the metadata propagation

Potential issues

1. slower start

[mhratson]

Make configurable?

[mhratson]

will revisit the implementation later…