Initial fixes for partition problem

[ehenneken]

This fixes the problem with the surprising results of the round() function in Python 3.8 versus Python 2.7. E.g. round(12.5) evaluates to 12 in the Python 3.8 version, while it evaluates to 13 in the Python 2.7 version. This causes different data going into the augment_graph_data function. With the implemented fix, the data remains the same in both versions up to the point where the network is partitioned. With the current fix, results are still different after the partitioning at https://github.com/adsabs/vis-services/blob/master/vis_services/lib/paper_network.py#L95. No idea how to fix this.

[marblestation]

This looks good! Thanks for investing time investigating this!