fix: SGD results now reproducible

[immu4989]

Why are these changes needed?

Summary

• Seed random_state on SGDEstimator so SGDClassifier / SGDRegressor produce deterministic results (uses config.get("random_seed", 10242048), matching the LinearSVC and ElasticNet fixes).
• Add "sgd" to test_reproducibility_of_classification_models and test_reproducibility_of_regression_models.
• Intentionally do not add "sgd" to the *_underlying_* variants: SGDEstimator wraps the sklearn model with a Normalizer preprocessing step that the test helper does not replicate, so a bare-sklearn refit cannot match the wrapper's CV result. A short comment in each list documents this.

Related issue number

Follows: #1369 (LGBM), #1374 (ElasticNet), #1376 (LinearSVC), #1364 (CatBoost).

Test plan

• pytest test/automl/test_classification.py -k "reproducibility and sgd" — passes
• pytest test/automl/test_regression.py -k "reproducibility and sgd" — passes

Checks

• I've used pre-commit to lint the changes in this PR (note the same in integrated in our CI checks).
• I've included any doc changes needed for https://microsoft.github.io/FLAML/. See https://microsoft.github.io/FLAML/docs/Contribute#documentation to build and test documentation locally.
• I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.

[immu4989]

@microsoft-github-policy-service agree

[Copilot]

Pull request overview

This PR aims to make FLAML's SGDEstimator reproducible, aligning it with recent reproducibility fixes for other estimators in flaml/automl. It updates the SGD wrapper to seed the underlying sklearn model and extends the existing reproducibility test suites to cover SGD behavior.

Changes:

• Seed SGDEstimator with a default random_state so SGDClassifier and SGDRegressor train deterministically.
• Add "sgd" to the wrapper-level reproducibility tests for classification and regression.
• Document why SGD is still excluded from the underlying-model parity tests due to its extra Normalizer preprocessing layer.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
flaml/automl/model.py	Adds deterministic seeding to SGDEstimator initialization.
test/automl/test_classification.py	Extends classification reproducibility coverage to SGD and documents the underlying-model test omission.
test/automl/test_regression.py	Extends regression reproducibility coverage to SGD and documents the underlying-model test omission.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

[thinkall]

LGTM! Thanks.

[immu4989]

Thanks for the review! All comments addressed and resolved. Ready to merge whenever convenient