Serializing with bnode graphs smushes them

Unlike trig, the JSON-LD serializer does not maintain distinct graphs when those graphs are identified using BNodes. The expectation should be to preserve the semantics as available, and keep the graphs distinct and identified (especially since the graph bnode itself can be used within a graph or another graph.

Here's an example in action:

>>> from rdflib import *
>>> g = ConjunctiveGraph()
>>> g1 = Graph(store=g.store, identifier=BNode())
>>> g1.add((URIRef("http://example.com/alice"),RDF.type,URIRef("http://example.com/Person")))
>>> g2 = Graph(store=g.store, identifier=BNode())
>>> g2.add((URIRef("http://example.com/bob"),RDF.type,URIRef("http://example.com/Person")))
>>> print(g.serialize(format="json-ld").decode('utf8'))
[
  {
    "@id": "http://example.com/bob",
    "@type": [
      "http://example.com/Person"
    ]
  },
  {
    "@id": "http://example.com/alice",
    "@type": [
      "http://example.com/Person"
    ]
  }
]
>>> g1.identifier
rdflib.term.BNode('N744ad2b3deb24f6b8c4a3798ca865a87')
>>> g2.identifier
rdflib.term.BNode('N3886a68ae1154b08ac7147347f5960d5')
>>> len(g1)
1
>>> len(g2)
1
>>> len(g)
2
>>> print(g.serialize(format="trig").decode('utf8'))

_:N3886a68ae1154b08ac7147347f5960d5 {
    <http://example.com/bob> a <http://example.com/Person> .
}

_:N744ad2b3deb24f6b8c4a3798ca865a87 {
    <http://example.com/alice> a <http://example.com/Person> .
}

I would expect the output of JSON-LD to be like this:

[
  {
    "@id" : "_:N3886a68ae1154b08ac7147347f5960d5",
    "@graph" : [
      {
        "@id": "http://example.com/bob",
        "@type": [
          "http://example.com/Person"
        ]
      }
    ]
  },
  {
    "@id" : "_:N744ad2b3deb24f6b8c4a3798ca865a87",
    "@graph" : [
      {
        "@id": "http://example.com/alice",
        "@type": [
          "http://example.com/Person"
        ]
      }
    ]
  }
]