Spike into event driven operation considerations

[iFergal]

Please see WebOfTrust#290

It would be very useful if we use server sent events to make KERIA more event driven for operation completion and notifications.

This spike is not to cover how to implement SSE between KERIA and Signify - this can be covered later. This spike is to investigate for each operation type, how we can make it more event driven - it will involve changes in keripy.

Currently, when you check for a long running operation completion, longrunning.py will check LMDB to see if all of the criteria are in place for an operation to be considered complete (or failed, etc). So it's basically a reflection of the database state.

If we want to make things more event driven, we need to track event completions in LMDB so that an SSE handler can inform the client of them. The client may want to read old messages too.

So this spike should investigate the effort of this in all operation types.

We should also consider if we can make better use of Failed states - for example, if an exn message is not valid, we should mark the operation as failed, instead of having a pending operation forever.