Investigate new heuristic "networking in setup.py"

[christophetd]

Based on:

e) Some rules are more effective than others: To understand why these tools flagged so many benign packages as malicious, we broke down the specific rules that
were triggered in the case of the PyPI malware checks.
Figure 3 shows the distribution of the alerts for each
rule in the setup.py files of the three datasets. We observed that metaprogramming_in_setup is the most
common rule triggered in the popular and random packages. However, malicious packages contain the highest percentage of networking_in_setup alerts. This indicates the indicators of a networking event could provide
a higher confidence of maliciousness.

from https://arxiv.org/pdf/2209.13288.pdf