PersistenceMechanism: who should close it?

[mauricioaniche]

Currently, the PersistenceMechanism is closed by repodriller, at the end of its execution.

@davisjam argues (and I agree) that it's a bit odd that the client opens it, but we close it. Read comments on #72 for more details.

We should only pay attention to not break the existing behavior as this would require people to upgrade all their existing studies.

[davisjam]

Agreed that this needs to be backwards-compatible.

I think the default behavior should be to close them, as this is probably what the user wants and certainly what the current behavior is.

I think it would be fine to have a way to mark some PersistenceMechanism's as "user managed" in terms of who should close them.