Upgrade scikit-learn

[eccabay]

Closes #4247

** New and improved **
Upgrades to the most recent versions of scikit-learn and imbalanced-learn, plus updating sktime and vowpal-wabbit.

** Previous **
The newest version of scikit-learn is incompatible with the version of imbalanced-learn that we're pinned to. Unfortunately, imbalanced-learn must remain pinned due to lack of nullable type handling, so we can't upgrade scikit-learn. Once imbalanced-learn fixes their nullable type issues, we should be able to unpin both.

Nonetheless, this includes changes to Decision Tree and Extra Trees estimators to remove a deprecated "auto" argument that will be fully removed once we do upgrade to 1.3.0.

[codecov]

Codecov Report

Merging #4248 (961e1fe) into main (5e158ed) will increase coverage by 0.1%.
The diff coverage is 100.0%.

@@           Coverage Diff           @@
##            main   #4248     +/-   ##
=======================================
+ Coverage   99.7%   99.7%   +0.1%     
=======================================
  Files        349     349             
  Lines      38309   38320     +11     
=======================================
+ Hits       38190   38201     +11     
  Misses       119     119             

Impacted Files	Coverage Δ
evalml/pipelines/component_graph.py	99.8% <ø> (ø)
...components/ensemble/stacked_ensemble_classifier.py	100.0% <ø> (ø)
...estimators/classifiers/decision_tree_classifier.py	100.0% <ø> (ø)
...components/estimators/classifiers/et_classifier.py	100.0% <ø> (ø)
...s/estimators/regressors/decision_tree_regressor.py	100.0% <ø> (ø)
...s/components/estimators/regressors/et_regressor.py	100.0% <ø> (ø)
evalml/pipelines/components/utils.py	96.5% <ø> (ø)
...ts/estimators/classifiers/kneighbors_classifier.py	100.0% <100.0%> (ø)
evalml/tests/component_tests/test_components.py	99.1% <100.0%> (ø)
...s/component_tests/test_decision_tree_classifier.py	100.0% <100.0%> (ø)
... and 6 more

[jeremyliweishih]

Thanks!

[eccabay]

In further investigation of the imbalanced-learn error that caused its pinning, I found a workaround since the error is just in one of our tests.

Pausing review on this PR until I can determine if this workaround allows me to unpin imbalanced-learn as well as upgrade scikit-learn instead of pinning it.

[eccabay]

Looks like this is blocked until conda-forge/sktime-feedstock#94 is merged

[fkiraly]

Looks like this is blocked until conda-forge/sktime-feedstock#94 is merged

No longer blocked :-)

[chukarsten]

Thanks for doing this!