Potential bug in the merge algorithm

In #48, I incorporated prefix synonyms from the Bioregistry. Since it links all of the many Wikidata CURIE prefix variants together (wd, wikidata, WD_Entity), it's surprising that there are still disconnected prefix expansions:

prefixmaps/src/prefixmaps/data/merged.csv

Line 4541 in b8a2bbd

merged,wd,http://www.wikidata.org/entity/,canonical,prefixcc

prefixmaps/src/prefixmaps/data/merged.csv

Line 4574 in b8a2bbd

merged,WIKIDATA,http://identifiers.org/wikidata/,canonical,bioregistry

I think this is a problem because of how the merging algorithm works. The issue might be that merge algorithm doesn't have a way to stitch two previously disjoint canonical CURIE/URI prefix records together when it's given synonyms (or it is doing it, but not getting the optimal results).

Maybe an alternative is to just fix #49 directly