Working towards 5.0.0

[laserkelvin]

This issue lays out the things I'd like to push out for the 5.0.0 release. All developments will be done in 5.0.0-dev, and when we're ready, PR to master.

New installation practices, CI/CD

• Use poetry for dependencies; migrate to pyproject.toml instead of the crazy setup.py currently. Poetry install #31
• Better categorization of dependencies.¹
• Migrate to Github actions for testing.

Better abstraction for spectral assignment

• Major refactor of AssignmentSession. Currently not very maintainable.
• Major refactor of analysis module; also not very maintainable.
• Rework of the line assignment procedure; ML models for maximum likelihood assignments.

Renewed interface to SPCAT

• Rotor abstractions. 2449441
• Interface allows user to add own terms
• Database for molecular parameters

Footnotes

1. The neural network stuff should be optional, as torch is massive. Similarly, testing suites for dev. ↩