ci: add workflow to deploy to pypi

[ap--]

This PR adds a workflow to deploy scikit_mol to pypi for commits that are tagged.

The way you use this is the following:

• develop your module on github and everytime you want to make a new release you create a tag

  # the current latest commit on main that should be released as a new version
  $ git tag v0.1.2
  $ git push origin v0.1.2

• this will trigger the ci and build a new version of the package and deploy to pypi
• I recommend having the tags adhere to semantic versioning vX.Y.Z

To set up the required tokens, goto pypi and sign up.
Goto your account settings and scroll to the API token section. Create a new API token with scope for everything (note: after your project was created you can replace the token with a token that has scope only for your project)
Copy the token and add it as a secret in your github repository (Settings > Secrets > Actions, --> Create New Repository Secret) The secret name should be PYPI_PASSWORD.

When the deploy pipeline after a tagged commit succeeds, your new version is available on pypi.

People can then just pip install scikit-mol

If anything is unclear, let me know!

[EBjerrum]

Thanks for the detailed descriptions, I've setup the pypi account and created the access token and added it for the repository. One question, do you have a good reference with instructions for the way that you have set up the repository?

[ap--]

The way I set it up is mostly a collection of different pieces from multiple best practices for how to setup these packages. It has little overhead and only uses the default pypa tools instead of relying on tools like flit or poetry.

Resources are here:

• https://www.pypa.io/en/latest/ the Python Packaging Authority
• https://packaging.python.org/en/latest/ checkout the guides
• https://github.com/pypa/setuptools_scm in my mind the most convenient versioning tool when working with git
• https://peps.python.org/pep-0660/ , Implement PEP 660 allowing both "strict" and "lax/loose" approaches pypa/setuptools#3265 setuptools supports editable installs without setup.py since version 64
• https://peps.python.org/pep-0517/ and https://peps.python.org/pep-0518/ regarding the move to pyproject.toml

And for running the tests in the CI, it's just running pytest together with coverage (for code coverage analysis) and the way the tests are setup via github actions uses caching for files you might have to always download for testing your code.

If you want I could make another PR to setup automatic source code linting / formatting, and maybe static type analysis via mypy in case you want to standardise your code in the way it's now commonly done in the Python community. Although I would probably not do that right now, because it will just increase the barrier for contributions without any imminent benefit for the currently small codebase.

Cheers,
Andreas 😃