Padelpy

No functional changes, but:

- Build/installation now uses `pyproject.toml` instead of the deprecated `setup.py`
- Removed Azure-based unit testing, using GitHub workflows and `pytest`
- New releases are automatically deployed to PyPI

As outlined in https://github.com/ecrl/padelpy/issues/46, if two parallel processes started their descriptor calculations at the same time, a single temporary file would be associated with each process. This fix adds a random number to the filename such that two processes started at the same time cannot share the same temporary file. Credit to pikakolendo02 for finding the error and supplying a fix.

The _padeldescriptor_ wrapper's _retainorder_ argument now defaults to _True_. It just makes more sense this way!

A few pull requests/features included in this update, thanks to all who contributed!

1.) Now cleans up orphaned CSV files if KeyboardInterrupt triggered
2.) Added option to control runtime per molecule
3.) Added option to specify the number of CPU threads used

This update allows users to generate descriptors/fingerprints from an SDF file.

- This update adds an additional argument to padelpy.wrapper.padeldescriptor, `headless`, True by default, to prevent the PaDEL-Descriptor splash screen from displaying on execution