Simplify scale

[flying-sheep]

scale_sparse has a lot of code that’s not needed.

This PR also fixes zappy compatibility for scale, not that anyone is using it as far as I know.

[codecov]

Codecov Report

Attention: Patch coverage is 86.66667% with 6 lines in your changes missing coverage. Please review.

Project coverage is 75.42%. Comparing base (a28fe88) to head (18a58d7).

Files with missing lines	Patch %	Lines
src/scanpy/preprocessing/_scale.py	86.66%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3351      +/-   ##
==========================================
- Coverage   75.44%   75.42%   -0.03%     
==========================================
  Files         113      113              
  Lines       13250    13233      -17     
==========================================
- Hits         9997     9981      -16     
+ Misses       3253     3252       -1     

Files with missing lines	Coverage Δ
src/scanpy/preprocessing/_scale.py	75.63% <86.66%> (-2.32%)	⬇️

[scverse-benchmark]

Benchmark changes

Change	Before [a28fe88] <1.11.0>	After [18a58d7]	Ratio	Benchmark (Parameter)
+	10.6±0.1ms	13.3±0.1ms	1.26	preprocessing_counts.FastSuite.time_calculate_qc_metrics('bmmc', 'counts-off-axis')
+	5.58±0.06ms	6.54±0.04ms	1.17	preprocessing_counts.FastSuite.time_calculate_qc_metrics('pbmc68k_reduced', 'counts')
-	60.5±1ms	50.8±8ms	0.84	preprocessing_counts.time_filter_genes('pbmc3k', 'counts-off-axis')
-	615±100μs	532±5μs	0.87	preprocessing_log.FastSuite.time_mean_var('bmmc', 'off-axis')
+	952M	1.31G	1.38	preprocessing_log.peakmem_scale('pbmc3k', 'off-axis')
+	1.15G	1.51G	1.32	preprocessing_log.peakmem_scale('pbmc3k', None)
-	753±10ms	481±1ms	0.64	preprocessing_log.time_scale('pbmc3k', 'off-axis')

Comparison: https://github.com/scverse/scanpy/compare/a28fe88ec08cc144189f6cc1497a5f067b96ce7b..18a58d77d59fd3e7912cd89990ffc779f9616d31
Last changed: Mon, 17 Feb 2025 18:50:47 +0000

More details: https://github.com/scverse/scanpy/pull/3351/checks?check_run_id=37353611695

[flying-sheep]

@Intron7 any idea what I did wrong? Why do the benchmarks say that it uses more memory?