Allocate an endpoint number.
Either allocates a new number for a Endpoint or reserves the endpoint's number. If the Endpoint already has a number but it is allocated to a different endpoint it is an error.
We must persist the assigned number and next endpoint number. We are fairly resilient to the small chance that persistence fails so we persist lazily and return synchronously.
Obtain the store for a single Endpoint.
These stores are cached internally by ID.
Manages all EndpointStores for a Node.
We eagerly load all available endpoint data from disk because this allows us to keep Endpoint initialization more synchronous. We can initialize most behaviors synchronously if their state is already in memory.
TODO - cleanup of storage for permanently removed endpoints