Fix random forest import with scikit-learn 1.9

[Aplietexe]

Reference Issues/PRs

See scikit-learn/scikit-learn#33948.
Related downstream report: automl/mighty_dr_example#1.

Description, Motivation and Context

scikit-learn 1.9 removed the private sklearn.tree._tree.DTYPE alias. SMAC's sklearn-based random forest currently imports that private symbol, so importing SMAC fails with scikit-learn 1.9:

ImportError: cannot import name 'DTYPE' from 'sklearn.tree._tree'

scikit-learn replaced this internal alias with direct np.float32 usage, so this PR does the same in SMAC.

Changes

• Remove the private from sklearn.tree._tree import DTYPE import.
• Use np.float32 directly when validating random forest prediction input.

Type Of Changes

What types of changes does your code introduce? Put an x in all the boxes that apply:

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

What is Missing?

Nothing known.

Checklist

Go over all the following points, and put an x in all the boxes that apply.
Should there be a good argument to differ, please provide a reasoned explanation.

• My change is based on the latest stage of the develop branch.
• My change required a change of the documentation, which has been done.
• I checked that the documentation can be build, visualizes everything as expected, and does not contain any warnings.
• I have added/adapted tests to cover my changes.
• The tests can be executed successfully.
• I have added a description of the changes to CHANGELOG.rst.

Any other comments?

None