Clarify role of blank nodes and IRIs in docmaps

[ships]

There are extensive issues with blank nodes variously documented. The most recent issue I am discovering is that since blank nodes are existential quantifiers, a SPARQL processor can't meaningfully modify a dataset with blank nodes; it is impossible to idempotently upload triples because every new blank node statement is assumed to refer to an unknown blank node (see spec) and same is true for deletes.

This issue is a placeholder for a future PR to clarify the role of blank nodes in docmaps.