Now we have persistent storage we want to make sure that we apply changesets in order. One way to ensure this is that each changeset commits to the previous one. This can be checked by the persistent storage backend before applying it.

Things to consider:

1. When we add new transactions to the sparse chain we may want to atomically remove conflicts in the mempool as dictated by the TxGraph see Remove conflicting transactions from mempool/unconfirmed #68. How do we make sure that these two things end up in the same changeset and avoiding having to return multiple changesets.