Extend benchmarks from basic tutorial

[flying-sheep]

• Closes Benchmarks: Add all from basic tutorial #3012
• Tests included or not required because:

• Release notes not necessary because:

I added a sparse dataset to cover more code paths in most benchmarks.

regress_out seems pretty slow with sparse data, maybe that should be tackled instead of hidden by disabling the benchmark.

[codecov]

Codecov Report

Attention: Patch coverage is 69.23077% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 76.27%. Comparing base (d2a5368) to head (69f9781).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3031      +/-   ##
==========================================
- Coverage   76.27%   76.27%   -0.01%     
==========================================
  Files         117      117              
  Lines       12799    12803       +4     
==========================================
+ Hits         9763     9766       +3     
- Misses       3036     3037       +1     

Files	Coverage Δ
scanpy/preprocessing/_highly_variable_genes.py	95.60% <100.00%> (ø)
scanpy/preprocessing/_scrublet/core.py	92.90% <100.00%> (+0.14%)	⬆️
scanpy/preprocessing/_simple.py	87.28% <100.00%> (ø)
scanpy/preprocessing/_scrublet/__init__.py	96.70% <0.00%> (-1.08%)	⬇️
scanpy/tools/_rank_genes_groups.py	94.33% <50.00%> (ø)

[scverse-benchmark]

Benchmark changes

Change	Before [d2a5368]	After [69f9781]	Ratio	Benchmark (Parameter)
-	1.24G	1.03G	0.83	preprocessing_counts.peakmem_scrublet('pbmc3k')
-	466M	366M	0.79	preprocessing_counts.peakmem_scrublet('pbmc68k_reduced')
-	7.53±0.3ms	6.76±0.08ms	0.9	preprocessing_counts.time_log1p('pbmc3k')
-	11.2±0.06s	3.22±0.08s	0.29	preprocessing_counts.time_scrublet('pbmc3k')

Comparison: https://github.com/scverse/scanpy/compare/d2a53680e312835b998077b4e25b254e98bcb5ba..69f9781b5ae3970d735c3b68879a791164d6a54d
Last changed: Mon, 13 May 2024 14:51:19 +0000

More details: https://github.com/scverse/scanpy/pull/3031/checks?check_run_id=24903996679

[flying-sheep]

Oof, half an hour is definitely too long. I wonder why it takes that long.

I can finish this up on friday

[ilan-gold]

@flying-sheep Not sure. Now that you mention it, the rapids benchmark also takes more time than I'd expect as well...The functions themselves are quite fast.