fix: XGBoostSklearnEstimator honors random_seed config

[immu4989]

Why are these changes needed?

Seed random_state on XGBoostSklearnEstimator.__init__ so xgb.XGBClassifier / xgb.XGBRegressor / xgb.XGBRanker produce deterministic results across runs and honor the FLAML-internal random_seed config. Uses the same defensive pattern as #1541 and #1546 , pop the random_seed key from self.params, and only set random_state when the caller has not already provided one.

XGBoostLimitDepthEstimator inherits from XGBoostSklearnEstimator, so this PR closes two of the remaining XGBoost items in tracking issue #1540 (the non-sklearn XGBoostEstimator path will be addressed in a follow up PR).

Before this change, the existing xgboost / xgb_limitdepth reproducibility tests passed only because the search space init values (subsample=1.0, colsample_bytree=1.0, colsample_bylevel=1.0) are themselves deterministic. A different dataset or max_iter that explored stochastic-subsample configs would not have been reproducible.

Related issue number

Tracking issue: #1540
Pattern reference: #1547 (RandomForest), #1541 (SGD), #1546 (LRL1)

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.