The idea is to isolate the PipelinedOplogApplier and have a mock OplogApplierBatcher feed it a stream of randomly generated oplog entry batches. We can feed the classic applier (OplogApplierImpl) the same stream as well, and compare the two to make sure the PipelinedOplogApplier's execution was correct.

We can also log a history of what the PipelinedOplogApplier did and check it against some rules; for example we can make sure that the dispatcher never dispatched a DDL while some workers were still working on a batch.