Fix krypton calculation

[tbody-cfs]

This PR addresses an issue where the radiated power $L_z$ was returning zero for Krypton.

The Issue

The atomic data files from OpenADAS for Krypton have slightly different electron temperatures at a single position

print(adas_datasets["effective_recombination"].dim_electron_temp.values[32])
print(adas_datasets["line_emission_from_excitation"].dim_electron_temp.values[32])

returns

7170.839465426523
7170.674352646886

When merging xr.DataArrays into an xr.Dataset, a silent xr.merge(..., join="outer") is used. This fills NaNs where data is "missing", and this goes on to corrupt the entire Lz calculation.

To fix this, we now interpolate before combining the rate coefficients into a single dataset, and then use xr.merge(..., join="exact") which will raise an error if the axes are not perfectly aligned.

In addition, this model slightly changes the plotting, to make it a bit easier to see when Mavrin data isn't defined.

Github Actions are extended to run poetry run radas for all species on new releases, or if manually triggered by a user.

[MishaVeldhoen]

Hey Tom, good fixes overall! My main worry is about the silent grid extrapolation due to clipping, and publishing of OpenADAS data.

I also find (more generally) some parts of the code are a bit hard to read because the JSON config has a compact structure, the details of which leak into the code itself. E.g., filename can be an int, or a list, which reads unnaturally. Could be fixed in a next PR where we fix up config parsing with dataclasses or pydantic models

[MishaVeldhoen]

LGTM!