roll back scipy to 1.12 to check pytest

[m-reuter]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 34.92%. Comparing base (98bc454) to head (888d3e2).
Report is 5 commits behind head on main.

❗ Current head 888d3e2 differs from pull request most recent head d213ac0. Consider uploading reports for the commit d213ac0 to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #60   +/-   ##
=======================================
  Coverage   34.92%   34.92%           
=======================================
  Files          14       14           
  Lines        2365     2365           
  Branches      297      297           
=======================================
  Hits          826      826           
  Misses       1455     1455           
  Partials       84       84           

Flag	Coverage Δ
unittests	34.92% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[m-reuter]

scipy 1.13 changed some behaviour in sparse matrices. For example when creating CSC matrices (as we do for the adjacency matrices) from a data array and indices, the former behaviour was that it would add data at the same index pair if it appeared multiple times. This seems to have changed in 1.13. and breaks our code.

[m-reuter]

I reported this as a potential bug:
scipy/scipy#20670

But if it is by design, we can work around this by initialising as COO matrix and then converting to CSC which sums up the entries again, also described here:

https://docs.scipy.org/doc/scipy/reference/sparse.html#example-2