[Refactor] Avoid Scala in pinot-kafka-xxx project

Right now the only places scala are needed are in `pinot-kafka-xxx` which forces a scala version of `2.10` and `2.11` respectively.
It introduces conflict at runtime when our own plugin loads another version of scala (e.g., `2.12`)

Based on code reading, the reason `kafka_2.10`/`kafka_2.11` are needed is because of `KafkaDataServerStartable`. That `StreamDataServerStartable` is not needed for our team but Kafka consumer is definitely needed.

One solution would be to move the `KafkaDataServerStartable` into a separated Plugin Jar, so we can avoid loading them. This way if my team needs scala I can be free to load whatever version we need.

Start by tracing the pinot-kafka-xxx dependencies and the KafkaDataServerStartable and StreamDataServerStartable entry points mentioned in the issue. Confirm why kafka_2.10 and kafka_2.11 are required, which component provides the Kafka consumer, and how a separate Plugin Jar would affect loading. Done means the unnecessary Scala versions no longer cause runtime conflicts while Kafka consumer support remains available.

Written by the indexing model from the issue text.