Package for distribution

[hugovk]

First step for #24.

This does several things to prepare for a PyPI release, like replacing setup.py with pyproject.toml, which is the new standard for Python packaging.

This defines several things:

• A build "backend". I put the popular Hatchling here, I use this one for most things, it's extensible and standards compliant: https://pypi.org/project/hatchling/
• Define the lowest Python version supported. This tells pip or uv to only install this package on that version or higher. When we drop 3.9 later in a future release, pip/uv will then install the last version that did support 3.9, so the end user still has a working install.
• A list of "Trove" classifiers. Choose whatever you want from the list at https://pypi.org/classifiers/. These are all optional and show up on PyPI, and can be useful to help people/programs find things, but I don't think they're used too much. I find it useful to list the Python versions the package has been tested on.
• The dependencies which will also be installed by pip or uv: NumPy and pandas.
• The optional dependencies, when you install with "extras". For example, like we now do on the CI: pip install .[tests]. (You may need quotes on macOS: pip install ".[tests]".) This replaces the requirements.txt file.
• Some URLs. You can add more here, like issues or documentation. See also https://packaging.python.org/en/latest/specifications/well-known-project-urls/#well-known-labels.

To try this out locally:

pip install build
python -m build

This will create a wheel and sdist in a dist directory, for example:

starpy-0.0.1-py3-none-any.whl
starpy-0.0.1.tar.gz

You can then install these, for example:

pip install starpy-0.0.1-py3-none-any.whl

And check the library works as expected.

After this, we could set up some a deploy workflow using Trusted Publishers.

[ArendPeter]

Looks great, I appreciate the update!