fix: use get_igraph_from_adjacency for graph construction

[flying-sheep]

• Closes Inconsistency in graphs used by sc.tl.leiden and sc.metrics.modularity leads to modularity mismatch #4099
• Tests included or not required because:

• Release notes not necessary because:

[codecov]

Codecov Report

❌ Patch coverage is 86.36364% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.61%. Comparing base (bb4e605) to head (30bc534).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/scanpy/neighbors/__init__.py	0.00%	1 Missing ⚠️
src/scanpy/tools/_paga.py	0.00%	1 Missing ⚠️
src/scanpy/tools/_utils.py	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4112      +/-   ##
==========================================
- Coverage   79.61%   79.61%   -0.01%     
==========================================
  Files         120      120              
  Lines       12782    12786       +4     
==========================================
+ Hits        10176    10179       +3     
- Misses       2606     2607       +1     

Flag	Coverage Δ
hatch-test.low-vers	78.85% <86.36%> (+<0.01%)	⬆️
hatch-test.pre	79.47% <86.36%> (+<0.01%)	⬆️

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

Files with missing lines	Coverage Δ
src/scanpy/_settings/presets.py	82.81% <100.00%> (ø)
src/scanpy/_utils/__init__.py	75.34% <100.00%> (+0.22%)	⬆️
src/scanpy/metrics/_metrics.py	97.14% <100.00%> (-0.05%)	⬇️
src/scanpy/plotting/_tools/paga.py	64.26% <100.00%> (ø)
src/scanpy/tools/_draw_graph.py	69.01% <100.00%> (ø)
src/scanpy/neighbors/__init__.py	82.04% <0.00%> (ø)
src/scanpy/tools/_paga.py	32.05% <0.00%> (ø)
src/scanpy/tools/_utils.py	86.76% <85.71%> (+0.19%)	⬆️

... and 1 file with indirect coverage changes

[ilan-gold]

Is this a replacement for #4113?

Also should we add some documentation for this new behavior to the presets?

cc @jpintar

[flying-sheep]

This one is backwards-compatible, #4113 needs some more work due to igraph not handling sparse arrays (only matrices).

[ilan-gold]

Just documentation stuff. I think we should be very clear what is going on, why, and what the plan is

[ilan-gold]

This is the change though right? Like users of modularity will start seeing a different number? And we're considering that a bug? Just want to be 100% clear. I would explain the rationale behind breaking this in the release note more

[flying-sheep]

Yes, and for scanpy 2.0 we’ll change it back again. Not 100% convinced with that back-and-forth, but changing it in leiden would mean to change its results or to create two graphs, one just for the modularity …

We could also not change this site and just merge the “in the future, do things like we do them here” part

[ilan-gold]

The path forward seems reasonable. I am fine with this. The above is a bug against our intention, to match leiden, and the way forward is to rectify an underlying bug with our intention, that leiden is using the wrong graph. The flip-flop is just the fact that we have two options in two places and made the wrong choice both times.