Add tests to ensure Espresso Reader remains in a consistent state if it suddenly crashes or is interrupted

[miltonjonat]

We should make sure that, if the reader crashes at some intermediate point (e.g., after inserting a new input but before updating the nonce), that everything remains in a consistent state.

This includes:

• Ensuring that restarting the Node+Reader resumes input processing from the last persisted point
• Ensuring that the database remains in a consistent state (i.e., that either nothing is committed or all changes are committed for a given input). This would involve checking if transaction begin/commit/rollback code is working as it should. One idea would be to mock those methods to see if they are being used appropriately.