fix: correct repr for RepeatedKFold and RepeatedStratifiedKFold (swev-id: scikit-learn__scikit-learn-14983)

[rowan-stein]

Summary

• Add repr to _RepeatedSplits to use the existing _build_repr helper, ensuring proper repr for RepeatedKFold and RepeatedStratifiedKFold.
• Add regression tests for default and non-default parameter combinations.

Linked Issue

• See Issue: Fix __repr__ for RepeatedKFold and RepeatedStratifiedKFold #67

Reproduction Steps

1. In Python, run:

from sklearn.model_selection import RepeatedKFold, RepeatedStratifiedKFold
repr(RepeatedKFold())
repr(RepeatedStratifiedKFold())

2. Before this fix, repr(...) returns the default object address strings.
3. After this fix, expected outputs:

RepeatedKFold(n_splits=5, n_repeats=10, random_state=None)
RepeatedStratifiedKFold(n_splits=5, n_repeats=10, random_state=None)

Observed failure and stack trace (pre-fix)

• Attempting to import and run from the source tree without building yields the following (when using the unbuilt source):

ImportError: No module named 'sklearn.__check_build._check_build'
It seems that scikit-learn has not been built correctly.

• The functional failure targeted by this PR is the incorrect repr of the repeated splitters, which manifests as object address strings rather than parameterized reprs.

Notes

• No CI run triggered manually per policy.
• Base branch: scikit-learn__scikit-learn-14983.

[noa-lucent]

Solid fix. Thanks for adding regression coverage.