Event when there's no metadata attached to an event, we'd still use the
constructor passing an empty metadata vector, resulting in an on-the-wire
representation with an empty trailing vector.
Particularly visible when just snooping events via websocat. There also
seems to be some bug with the timestamp -1 handling.
The previous approach ignored the fact that nested / inner values might
also be Broker::Data values. I'm not super sure about the validity of
the test, because it's essentially demonstrating any-nesting, but
it's not leading to extra Broker::Data encoding.
This class is a parameter of virtual methods of the Backend API for users
to implement and also a parameter to the HookPublishEvent() API. Seems it
shouldn't be in detail and instead we should own it.
Alternatively, could mark the cluster APIs as not-stable-yet, but I
think we can move forward and make it non-detail for 8.0.
The broker serializer leverages the existing data_to_val() function.
During unserialization, if the destination type is any, the logic
simply wraps the broker::data value into a Broker::Data record.
Therefore, events with any parameters are currently exposed to
the Broker::Data type.
There is a bigger issue in that re-publishing such Broker::Data
instances would encode them as a normal record. Explicitly prevent
this by serializing the contained data value directly instead, similar
to what Broker already did when publishing a record.
