Roadmap proposals

[maresb]

Once we get some more time, here's a list of things on my mind that we could try:

• Fix PyPI release (hopefully just a capitalization issue)
• Rename master to main
• Drop support for older Python versions (Scientific Python recommends dropping support for anything less than 3.11, although I find that a bit extreme.)
• Add formatter
• Add linter
• Add type checks (or not?)
• Fail harder on invalid input
• Depend lazily on numba
• Add conda-forge installation instructions and badge
• Fix relative PyPI links via hatch-fancy-pypi-readme
• Benchmarking
  • Vectorize Newton iteration
  • Add more corrections to $\partial S / \partial T$ approximation to accelerate convergence
  • Use LUT
  • Try tightening convergence thresholds
  • Write a rust implementation
• Stress tests (check for potential failures)