Recover 15 superconductors lost to the std filter, and fix the tc aggregation (relates to #3)

[dl-eigenart]

Follow-up to #3, as suggested by @pfriederich.

What this recovers

Fifteen materials that are currently dropped from 3DSC entirely come back, among them four that most T_c work would consider standard:

norm_formula	compound	T_c
C25N25Nb50	NbC₀.₅N₀.₅	17.9 K
C10N40Nb50	NbC₀.₂N₀.₈	17.6 K
Al25Nb75	Nb₃Al	17.5 K
Si25V75	V₃Si	16.77 K
N50Nb50	NbN	13.54 K
C48.98Nb51.02	NbC	9.8 K
C89.552H4.478K4.478N1.493	K-doped carbon	28.0 K
Cu13.571Fe0.714La25.714Sr2.857O57.143	La-Sr-Cu-O, Fe-doped	30.0 K

• 7 more |   | 7.3–25 K

Nothing is removed and no class shifts. The 19 corrected values are the low-T_c cases from #3 that originate here, e.g. Ir50Te50 1.5 → 3.0 K and Bi50Sb50 1.47 → 2.94 K.

Verification

The check script is attached. The formula normalisation it uses was rebuilt for the test and reproduces norm_formula_sc from 3DSC_MP.csv for 5 726 of 5 773 entries (99.19 %); the remainder are sub-permille rounding differences from pymatgen parsing. Running inside the repository removes even those, since the original function applies.

Reach

This fixes every case that originates in this repository. Of the 139 entries listed in #3, 34 are of that kind. For the remaining ~105 the halved value is already present in Supercon_data_by_2018_Stanev.csv — it predates this pipeline and would need re-aggregation from MDR SuperCon, which is a separate question and not one a patch here can answer.

Also

The comment and print at that spot said "median" while the code computed mean; corrected alongside.

Best wishes, Daniel Leonforte

z12_patch_wirkung.csv
verify_patch.py

[dl-eigenart]

Reference for this change, in case it is useful for the commit message or a
future release note:

https://doi.org/10.5281/zenodo.22235399

Section 4 of the deposit documents the measured effect of this patch on the
repository's own input file — the fifteen restored compounds with their
transition temperatures, the nineteen corrected values, and the verification
that no entry is lost and the non-superconductor class is unchanged. The
verification script is included there as well.

Happy to amend the patch in any direction you prefer.