Support different karafka/waterdrop configurations per topic

[Drowze]

This patch depends on the other two patches that have pull requests open:

• Add WaterDrop integration (i.e.: trace messages produced by Karafka) DataDog/dd-trace-rb#4874
• Fix losing original karafka trace after iterating through the messages with distributed_tracing on DataDog/dd-trace-rb#4876

I'll open a new PR to integrate this patch as well once the previous patches get reviewed and merged upstream.

---

What does this PR do?

Adds support for multiple Karafka/Waterdrop configurations, based on the topic name.

Motivation:

Sometimes it makes sense to have distributed_tracing on for some topics but not for others, example scenario:

• in most topics, consumers process iterate through the messages and process them as usual - distributed_tracing makes sense here
• in some topics, consumers "pre-process" the messages into smaller batches (likely their own objects, not a Karafka's MessageBatch), so they can be processed more efficiently later - distributed_tracing makes little sense here (and can actually make the spans harder to follow). In this scenario, the application should probably handle the tracing continuation itself (if so desired).

In the above scenario, a configuration could look like:

Datadog.configure do |c|
  c.tracing.instrument :karafka, distributed_tracing: true
  c.tracing.instrument :karafka, describes: "critical_topic", distributed_tracing: false
end