add ensemble_bootstrap; add backend argument passed to joblib; add base_model_method argument to predict method

[chenyangkang]

1. ensemble_bootstrap argument: Defaults to False. if True, the data will be bootstrapped once for each ensemble. In this case users can generate ensemble-level uncertainty, accounting for variance in data.
2. joblib_backend argument: Defaults 'loky'. Other available arguments include 'threading' ('multiprocessing' will not work with generator as the return). Sometimes only threading may work on certain systems.
3. base_model_method argument: defaults to None. If None, predict or predict_proba will be used depending on the tasks. This argument is handy if you have a custom base model class that has a special prediction function. Notice that dummy model will still predict 0, so the ensemble-aggregated result is still an average of zeros and your special prediction function output. Therefore, it may only make sense if your special prediction function predicts 0 as the absense/control value. Defaults to None.

Only updated for AdaSTEM and STEM, not for SphereAdaSTEM.

[codecov]

Codecov Report

Attention: Patch coverage is 97.67442% with 4 lines in your changes missing coverage. Please review.

Project coverage is 90.76%. Comparing base (61efb24) to head (6baf66a).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
tests/test_model_custom_base_model_method.py	93.54%	2 Missing ⚠️
stemflow/model/AdaSTEM.py	96.55%	1 Missing ⚠️
stemflow/model/static_func_AdaSTEM.py	92.30%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #71      +/-   ##
==========================================
+ Coverage   90.36%   90.76%   +0.40%     
==========================================
  Files          35       38       +3     
  Lines        2594     2751     +157     
==========================================
+ Hits         2344     2497     +153     
- Misses        250      254       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[chenyangkang]

When you resample/bootstrap, or change any structure of the dataframe in a subprocess/thread or another python worker, a new copy of the data will be stored in the variable, doubling the memory cost. Resampling only the index will solve it.