Optimize localization by writing updates to relationships and nodes to disk first, then sorting and re-reading them in smaller chunks, writing them to Neo4j. Parallelizes localization by sequence node as well.
Will replace the existing localize_nodes(), used by build_pangenome. Checked code with @jonkh004 today. Also sanity-checked output by comparing build_pangenome output of master and this branch's version of PanTools on two small yeast data sets.