feat: pp.scale allow selecting correction

[AxelMcKenna]

Follow-up to #4196.

This adds a selectable correction parameter to sc.pp.scale. The default remains correction=1, so existing behaviour is unchanged.

The parameter is propagated through array, masked-array, sparse, and AnnData scaling paths. The documentation now explains how the variance divisor is calculated.

• Tests included for correction=0 and correction=1, covering dense, CSR, and CSC inputs; direct arrays and AnnData; masked and unmasked observations; and centered and non-centered scaling.

• Tests included or not required because:

• Release notes not necessary because:

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.93%. Comparing base (8673b4d) to head (9af49aa).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4220   +/-   ##
=======================================
  Coverage   79.92%   79.93%           
=======================================
  Files         121      121           
  Lines       12958    12958           
=======================================
+ Hits        10357    10358    +1     
+ Misses       2601     2600    -1     

Flag	Coverage Δ
hatch-test.low-vers	79.05% <100.00%> (+<0.01%)	⬆️
hatch-test.pre	79.89% <100.00%> (+<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/preprocessing/_scale.py	89.58% <100.00%> (ø)

... and 1 file with indirect coverage changes

[flying-sheep]

Please change to the standard name, correction: https://data-apis.org/array-api/latest/API_specification/generated/array_api.std.html