Create PyPI packages dataset

[nightlark]

Question: how do we handle PyPI packages that have native libraries? e.g. numpy

Script to create PyPI dataset (converted into sqlite database with name of package on PyPI and "import names") is pushed to a branch -- need to figure out how to process the data.

[nightlark]

• Any normalization/lowercase of import names?
  • Maybe already in the form they should be for importing? Mostly check to see if any platforms can import with arbitrary case while another one can't
  • casefold vs lowercase: German ẞ => ss?
• Submodules that get installed?
  • odoo
• Binary files in wheels?
  • More interesting for scanning a specific wheel/Python package installation, rather than capturing in dataset

[nightlark]

1/23 - using lowercase probably works; Windows treats the corner cases like German characters as separate paths (Windows is not doing casefolding for directory names)