In theory, Delete/Put could fail. If it does, we'll return the buffers
to the pool before we're really done with them.
In practice, this is almost certainly not an issue as badger shouldn't
_use_ the buffer unless we flush. But I feel slightly safer this way.
This paves the way for better object lifetime management.
Concretely, it makes it possible to:
- have different stores backing chain and state data.
- having the same datastore library, but using different parameters.
- attach different caching layers/policies to each class of data, e.g.
sizing caches differently.
- specifying different retention policies for chain and state data.
This separation is important because:
- access patterns/frequency of chain and state data are different.
- state is derivable from chain, so one could never expunge the chain
store, and only retain state objects reachable from the last finality
in the state store.
