[FEATURE] Improve prune dump and integration test

[imyousuf]

Is your feature request related to a problem? Please describe.
The first iteration of prune writes the messages twice, once within the message and then within the job object. This will cause unnecessary space consumption. So we will want to control what is written to the JSONL files. Also the integration test does not verify what is being exported, we should have some assertion.

Describe the solution you'd like
Message payload should be exported exactly once.

Describe alternatives you've considered
None considered

Additional context
This issue is a branch of #154