monarch_semsim doesn't handle identical ID phenotypes well

[oneilsh]

This example should produce two pairs of nodes best matching each other in both directions:

phenos1 <- monarch_engine() |> 
	fetch_nodes(query_ids = c("HP:0001763", "HP:0006316")) |>
	activate(nodes) |>
	mutate(source = "phenos1")

phenos2 <- monarch_engine() |> 
	fetch_nodes(query_ids = c("HP:0001763", "HP:0000678")) |>
	activate(nodes) |>
	mutate(source = "phenos2")


semsim <- monarch_semsim(phenos1, 
                         phenos2, 
                         include_reverse = TRUE, 
                         keep_unmatched = TRUE)

plot(semsim, node_color = source)

However, the identical node present in both sets has a self loop in phenos1 and no connection to phenos2.

Diagnosis: this comes down to how graph joins work in tidygraph (which is natural, over all matching columns) when the two node sets disagree in one of the columns. I think the desired behavior is right and desired with respect to that, we're just not keeping the original node sets properly in creating the join.