Restart failed tasks in Kafka Connect up to a configurable max-tries

Issue

Currently, Kafka Connect just reports failed tasks on REST API, with the error. Users are expected to monitor the status and restart individual connectors if there is transient errors. Unfortunately these are common for database connectors, e.g. transient connection error, flip of DNS, database downtime, etc. Kafka Connect silently failing due to these scenarios would lead to stale data downstream.

Proposal

Kafka Connect should be able to restart failed tasks automatically, up to a configurable max-tries.

Prior arts

• https://github.com/strimzi/proposals/blob/main/007-restarting-kafka-connect-connectors-and-tasks.md 
• https://docs.aiven.io/docs/products/kafka/kafka-connect/howto/enable-automatic-restart