This is just clustering with weights if I understood it right, so Louvain Method might worth checking out, but it could be a bit too heavy weight since you are looking at the interaction between less than a hundred countries.
One quick idea that is simpler: could you maybe model the votes as links, and assign them different strength, then maybe use a Force directed graph drawing algorithm to layout the while graph, and finally does a clustering? Of course, the final visualization should be rather easy to cluster just by looking at the result.
For this method, you could also generate a graph file, then use a tool like Gephi to do it.
Also, here's a related post: Are there implementations of algorithms for community detection in graphs?