🩹 Fix bad ports (2)

[s-weigand]

Follow up PR for #102 with refinements based on comparison errors and the work in #103

Change summary

• 🩹 Use same data in ex_doas_beta as in script
• ♻️ Rename notebooks for easier identification in CI download bundle
• 🚇 Compress results after running notebook
• 🩹 Refine ex_spectral_constraints to run with and without penalties
• 🩹 Use const parameters instead of number for activation
• 🩹 Fix ex_spectral_constraints result data file save location

Checklist

• ✔️ Passing the tests (mandatory for all PR's)

[sourcery-ai]

🧙 Sourcery is reviewing your pull request!

---

Tips

• Trigger a new Sourcery review by commenting @sourcery-ai review on the pull request.
• You can change your review settings at any time by accessing your dashboard:
  • Enable or disable the Sourcery-generated pull request summary or reviewer's guide;
  • Change the review language;
• You can always contact us if you have any questions or feedback.

[sourcery-ai]

We've reviewed this pull request using the Sourcery rules engine. If you would also like our AI-powered code review then let us know.

[jsnel]

Found what the cause of the CI fail (in sim-3d-weight) here was.

The issues is/was that the parameter values for (some of the element of) the activation would go wild. Specifically with non-negative=true and optimization method=lm.

Ultimatelly this would lead to:
Parameter(label='irf.width2', value=0.0, non_negative=True)
and other wierd values such as:
Parameter(label='irf.center3', value=5.695921809916089e+60, non_negative=True)

The value of 0.0 for irf.width2 causes a division by zero in
calculate_matrix_gaussian_activation_on_index specifically in beta = (t_n - center) / (width * SQRT2)
(at least when compiled in numba, in pure python it only results in a runtime warning)

The exact 0 value is most likely because the original estimate (per the optimizerI for this value is -96, which with {"non-negative": True} becomes 0.

And that this value is going wild, is because scaling is not working properly 😄 because of glotaran/pyglotaran#1463

So we find ourself in a proper chicken-egg situation here, so for the CI to pass here, I must first merge in glotaran/pyglotaran#1461

[jsnel]

We went from
214 failed, 363 passed, 49 warnings here
to
203 failed, 374 passed, 49 warnings here
to
159 failed, 418 passed, 49 warnings now here

and all tests are running again. ^^

[jsnel]

Reviewer and (extensively) tested ok. Large step in validation parity!