Enables threadpoolctl to control number of numpy threads

[franchuterivera]

addresses #924

One can see with threadpool_info() that the number of threads is set to 1 (this at the end of abstract evaluator, for debug purposes):

[CRITICAL] [2021-06-18 09:10:32,054:Client-TrainEvaluator(5):breast_cancer] threadpool_info()=[{'filepath': '/home/chico/anaconda3/envs/py38/lib/python3.8/site-packages/scikit_learn.libs/libgomp-f7e03b3e.so.1.0.0', 'prefix': 'libgomp', 'user_api': 'openmp', 'internal_api': 'openmp', 'version': None, 'num_threads': 1}, {'filepath': '/home/chico/anaconda3/envs/py38/lib/python3.8/site-packages/numpy.libs/libopenblasp-r0-5bebc122.3.13.dev.so', 'prefix': 'libopenblas', 'user_api': 'blas', 'internal_api': 'openblas', 'version': '0.3.13.dev', 'num_threads': 1, 'threading_layer': 'pthreads'}, {'filepath': '/home/chico/anaconda3/envs/py38/lib/python3.8/site-packages/scipy.libs/libopenblasp-r0-085ca80a.3.9.so', 'prefix': 'libopenblas', 'user_api': 'blas', 'internal_api': 'openblas', 'version': '0.3.9', 'num_threads': 1, 'threading_layer': 'pthreads'}]

[codecov]

Codecov Report

Merging #1161 (4010f80) into development (ba3f76f) will increase coverage by 0.23%.
The diff coverage is 100.00%.

@@               Coverage Diff               @@
##           development    #1161      +/-   ##
===============================================
+ Coverage        85.81%   86.05%   +0.23%     
===============================================
  Files              137      138       +1     
  Lines            10625    10663      +38     
===============================================
+ Hits              9118     9176      +58     
+ Misses            1507     1487      -20     

Impacted Files	Coverage Δ
autosklearn/evaluation/abstract_evaluator.py	90.15% <100.00%> (+0.11%)	⬆️
...tosklearn/metalearning/metafeatures/metafeature.py	62.13% <0.00%> (-14.57%)	⬇️
...eline/components/feature_preprocessing/fast_ica.py	91.30% <0.00%> (-6.53%)	⬇️
...onents/data_preprocessing/rescaling/standardize.py	95.23% <0.00%> (-4.77%)	⬇️
...ents/data_preprocessing/rescaling/robust_scaler.py	96.87% <0.00%> (-3.13%)	⬇️
...data_preprocessing/rescaling/abstract_rescaling.py	91.30% <0.00%> (-1.01%)	⬇️
...omponents/data_preprocessing/data_preprocessing.py	90.09% <0.00%> (-0.23%)	⬇️
...osklearn/metalearning/metafeatures/metafeatures.py	94.59% <0.00%> (-0.12%)	⬇️
autosklearn/estimators.py	93.38% <0.00%> (-0.10%)	⬇️
autosklearn/util/logging_.py	62.68% <0.00%> (-0.02%)	⬇️
... and 34 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ba3f76f...4010f80. Read the comment docs.

[mfeurer]

This looks good, thanks a lot.

However, it would be great to document this, for example as also suggested in #1009

It would be especially helpful to remind the user that this does not apply at prediction time.

[franchuterivera]

What do you think if we document it like this recent push suggest?

[mfeurer]

Thanks, the documentation looks good to me. But unfortunately, there's currently a documentation build error. What do you think about additionally linking https://scikit-learn.org/stable/computing/parallelism.html ?