perf: welford's algorithm for mean-var aggregation

[ilan-gold]

From discussions with @zboldyga.

This should then in theory be reused with #4143 instead of its custom moments calculation

• Closes #
• Tests included or not required because:

• Release notes not necessary because:

[ilan-gold]

We should make these nogil or provide an option for fau to provide nogil njit

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.60%. Comparing base (0ac3337) to head (48230af).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4147      +/-   ##
==========================================
- Coverage   79.61%   79.60%   -0.02%     
==========================================
  Files         120      120              
  Lines       12786    12780       -6     
==========================================
- Hits        10180    10173       -7     
- Misses       2606     2607       +1     

Flag	Coverage Δ
hatch-test.low-vers	78.84% <100.00%> (-0.01%)	⬇️
hatch-test.pre	79.43% <100.00%> (-0.05%)	⬇️

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

Files with missing lines	Coverage Δ
src/scanpy/get/_aggregated.py	92.66% <100.00%> (-0.65%)	⬇️
src/scanpy/get/_kernels.py	100.00% <ø> (ø)