Major package rewrite: structure, formatting, testing etc...

[martinjohndyer]

This is the big package rewrite and tidy up.

Some highlights are:

• Reworking the structure, for instance moving most functions into larger modules, having separate data and script submodules.
  • There are now only a handful of more general modules:
    • snf_simulations.data: handles loading data files and spectra
    • snf_simulations.spec: functions for creating spectra
    • snf_simulations.cask: contains the function for calculating the total spectrum of an SNF cask
    • snf_simulations.physics: various functions to do with nuclear physics calculations
  • The command_line.py script is now package entry point snf-sim, which should be available on the command line when the package is installed.
  • Most hardcoded data, like isotope fractions for the different reactors and atomic properties, are now stored in CSV files. Ultimately these should be replaced by a mix of input files (Add input configuration options #17), automated requests (Verify antineutrino data files #16) and the mendeleev package (Clarify atomic/molar masses and find better source #21).
• Replacing setup.py with a new pyproject.toml including package metadata (see Improve package metadata #8).
• Full code formatting and linting with ruff, including a GitHub Action workflow to check future pull requests (to run locally use ruff format & ruff check).
• Added full docstrings to each function, including type hints (Improve internal documentation #10), paving the way for full documentation (Create external documentation #11)
  • These are checked using ty (ty check), but there are a lot of outstanding errors due to ROOT. Reading around, I think [ty] Add replace-imports-with-any option astral-sh/ruff#23122 is what we'd need, but if we're aiming to drop ROOT anyway (Look at removing the dependency on ROOT #14) it's not an issue.
• (Almost) complete testing suite (Add testing and CI #6).
  • I say almost, because the code coverage is actually down from Add initial testing functions #18 at a somewhat depressing 59%. That's actually very heavily weighted towards the command_line.py script, which can't be tested since its main purpose is to produce ROOT windows. Coverage of the rest of the package is 95%+, with the exceptions being somewhat awkward cases like data files being unavailable (which again should be reworked in the future).
  • Automated testing is also not currently enabled, I tried to make a GitHub Action to run Pytest but again installing ROOT is a major hurdle.

There were also a handful of new bugs I came across while putting this together, which are all now fixed:

• Secondary decays are cropped one line too early #19 (the data selection isn't hardcoded any more, so no incorrect cropping)
• Standardise how max_E is determined for scaled histograms #20 (the spectra are now scaled automatically to the maximum energy bin)
• Checking mass of single vs multiple cask cases #22 (the cask mass is corrected to 10 tonnes for a single cask, and 10 * 10 for the multiple sets of casks)

Some things I didn't get around to doing:

• I think it would be more usable to make the module more classed base, e.g. with a Spectrum class for the basic unit and then a Cask class which could combine spectra (you can see the idea in how the modules are named). I did start working on this, but a lot of it would be built around the ROOT classes, especially ROOT.TH1D and ROOT.TList. So rather than doing a load of work now that would end up being removed, I'll wait until Look at removing the dependency on ROOT #14 which I think deserves its own PR (and is probably the next task anyway).
• Another thought was to replace the command_line.py script with something like a Jupyter Notebook, since it's really more of a demo of the various features of the package. But again that doesn't really work with ROOT plots. So the snf-sim script works for now, and when ROOT is replaced that can be reworked into a more general command line interface (which might also include launching the actual GUI too).
• I also didn't touch the README yet, since it will need rewriting along with the full documentation and that should wait until the API is finalised.