[DBZ-PGYB] Decouple connector to not share the same PostgresPartition to sourcePartition mapping across tasks

[yugabyte-ci]

Jira Link: DB-13591

With the introduction of the parallel snapshot model, we can have multiple tasks when the snapshot mode is set to parallel. This introduces a problem at the underlying layer when the connector stores the sourceInfo for its partitions i.e. PostgresPartition objects in Kafka.

The PostgresPartition is identified by a map which has a structure {"server", topicPrefix} - currently this is the same for all the PostgresPartition objects which are created by the tasks when snapshot.mode is parallel and hence they all end up referring to the same source partition in the Kafka topic. Subsequently, what happens is that (assume that we have 2 tasks i.e. 0 and 1):

One task (task_0) completes the snapshot while the other is yet to start.
a. After completion, task_0 updates the sourceInfo saying that its snapshot is completed.
b. When task_1 starts up, it reads the same sourceInfo object and concludes that the snapshot is completed so it skips its snapshot.

The above situation will cause a data loss since task_1 will never actually take a snapshot.

[vaibhav-yb]

This PR adds a short term fix by adding the task ID to the PostgresPartition which helps in uniquely identifying a Kafka source partition. However, the longer term fix would be to add the hash ranges to the partition as they will not change while the task ID can change and point to a different hash range if there’s any change to the distribution logic in the future.

yugabyte/debezium#163