Even smarter merge for partial metadata

[apirogov]

Specifically when using JSON LD, there is a difference between the semantic identity of objects and the available information.
Multiple pieces of information can refer to the same entity, witnessed by equal @id.

Currently, the "smart merge" of partial metadata would not combine two instances partial instances with same @id together when they appear in a list or set, because structurally they are different.

A general solution would be to define __identity_fields__ at MetadataSchema to be a tuple that, if present, defines the "semantic identity". In the case of JSON-LD-based schemas, we would simply set __identity_fields__ = ("id_",)

The smart merge then could be extended so that it would inspect objects in lists or sets, and instead of doing concat/set union as usual, for suitable objects (MetadataSchema with set __identity_fields__) it would instead combine them.

When this new special field is not set, each instance would be treated as a separate identity and never be merged with others.

On the other hand, it is not clear how useful this would be, as this presupposes that usually the @id for some given schema is something consistent either by force or convenion (e.g. ORCID of an author/Person, thus allowing to combine partial author information).