Sheshuk/fix model notebooks

[Sheshuk]

Closes #392 , #223

Trying to fix the remaining issues with building the notebooks:

• Errors in most PinchedModel-derived model classes, because the flux.Container failed to initialize with a scalar time (shape mismatch)
• Errors in Fornax_* models - caused by extra parameters interpolation, phi and theta. We need to discuss how to handle this in Make signature for SupernovaModel.get_*_spectra uniform for all models/subclasses #223
• Error in Fornax_2019 model: Fornax_2019 initial spectra unit is different from other models #403
• Errors in SNOwGLoBES_models nb - it needs updating, it uses path to already created tar.gz file.
• Errors in AnalyticFluence - it's incompatible with our ccsn_loaders.

Also, currently building the notebooks in the docs is not very informative: if one fails, we don't get any additional information. I'm adding a separate step, calling the pytest on notebooks, so we can see exctly which ones of them are failed, instead of exit on first failure.

[JostMigenda]

[ ] Errors in AnalyticFluence - it's incompatible with our ccsn_loaders.

In #404, @jpkneller suggested (among other things) moving that class from ccsn.py to base.py. So we could leave this particular issue for that PR and don’t have to deal with it here.

[JostMigenda]

I’d like to be very strict here, because any errors that slip through will make it to the published documentation.

[Sheshuk]

It is still strict: I've added the pytest --nbval step, so we're still having the notebooks checked, but it provides more information: instead of just failing on the first failed notebook during the documentation build, it will run each of them as test and give the output.