[staging] 🐛 Add back per element A-/K-Matrices to result dataset

[s-weigand]

Currently staging saves the combined A-/K-Matrices for all decay elements.
If we need those is something to be decided by @ism200 .

However, what we need for validations are data variables on a per-element basis with the following name pattern

• a_matrix_<element_name>
• k_matrix_<element_name>
• k_matrix_reduced_<element_name>

To get rid of the spammy xarray warning:

UserWarning: Duplicate dimension names present: dimensions {'species'} appear more than once in dims=('species', 'species').

Please use ('to_species', 'from_species') notation.

Example outputs

main

staging

[jsnel]

As discussed, please add the <element_name> suffix at the end.
a_matrix_<element_name>
k_matrix_<element_name>
rather than a element prefix.

Note element is new name for megacomplexes.

--

And in another example:
decay_associated_amplitudes_<element_name>
e.g.:
decay_associated_amplitudes_mc1
decay_associated_amplitudes_mc2

A special case is species_associated_amplitudes which we discussed is best to also save with the suffix, which before was only saved at a global level. But with the suffix in place it can then be consolidated later into the global amplitudes (spectra) for plotting purposes (in the extras).

[jsnel]

After experiencing the results in #1534 we decided on a small pivot, we're unflattening the results.

Something along the lines:

result
- data: dict (of Datasets)
- elements: dict of dicts (of Datasets of ElementResults) | None
  # using dataset attributes to 'type' ElementResults
- activations: ? (tbd)
- optimization properties : varied # can remain at top level, be at .optimization

Then, in the case of a kinetic element, at the level of
result.elements["dataset"]["element_name"]

You just have

- species_associated_amplitudes
- kinetic_associated_amplitudes
- species_associated_concentrations
- k_matrix | k_matrix_full (or reduced, you know)
- a_matrix