expand() doesn't connect duplicated nodes

[oneilsh]

If given a query graph with duplicate nodes by id (for example, two instances of HP:0100775, but with a source_disease attribute added with values "eds" and "marfan"), then expand() will only expand edges from one of them; it should do all.

The reason is because expand() (for both file engines and neo4j engines) uses tidygraph's graph_join() to join the query graph with the fetched result graph, and graph_join() does not add such edges (nor will it even when joining graphs with exactly duplicate nodes).

To fix this I propose adding a kg_join() function as a more-expansive version of tidygraph's graph_join() which replicates any edges (according to id) across all node pairs in the node and edge data, and using that in expand(). This would be a useful function anyway when working with more advanced use cases.