Remove the other JIRA issue edge while deduplicating

Since we only deduplicate JIRA issue edges when we build undirected networks,
it does not matter which of the two duplicates we remove. Some tests can be
adjusted more easily to the JIRA edge deduplication, by swapping which of the
edges get removed.

Signed-off-by: Maximilian Löffler <s8maloef@stud.uni-saarland.de>

util-networks.R

@@ -520,15 +520,15 @@ NetworkBuilder = R6::R6Class("NetworkBuilder",
                     ## if a match is found, remove 'add.link' and its corresponding referenced_by event
                     if (nrow(match) > 0) {
                         add.links = add.links[!(
-                            add.links$issue.id == add.link$issue.id &
-                            add.links$event.info.1 == add.link$event.info.1 &
-                            add.links$date == add.link$date &
+                            add.links$issue.id == match$issue.id &
+                            add.links$event.info.1 == match$event.info.1 &
+                            add.links$date == match$date &
                             add.links$author.name == add.link$author.name), ]
                         referenced.bys = referenced.bys[!(
-                            referenced.bys$issue.id == match$issue.id &
-                            referenced.bys$event.info.1 == match$event.info.1 &
-                            referenced.bys$date == match$date &
-                            referenced.bys$author.name == match$author.name), ]
+                            referenced.bys$issue.id == add.link$issue.id &
+                            referenced.bys$event.info.1 == add.link$event.info.1 &
+                            referenced.bys$date == add.link$date &
+                            referenced.bys$author.name == add.link$author.name), ]
                         matched = c(match, add.link)
                     }
                 }