swev-id: scikit-learn__scikit-learn-25931: Fix IsolationForest feature names warning during fit when contamination != 'auto'

[casey-brooks]

Reproduction (before fix)

```bash
PYTHONWARNINGS=always python - <<'PY'
import pandas as pd
from sklearn.ensemble import IsolationForest

X = pd.DataFrame({"a": [-1.1, 0.3, 0.5, 100]})
IsolationForest(random_state=0, contamination=0.05).fit(X)
PY
```

Observed warning
```text
/workspace/scikit-learn/.venv/lib/python3.11/site-packages/pandas/core/algorithms.py:1743: DeprecationWarning: is_sparse is deprecated and will be removed in a future version. Check isinstance(dtype, pd.SparseDtype) instead.
return lib.map_infer(values, mapper, convert=convert)
/workspace/scikit-learn/sklearn/utils/validation.py:605: DeprecationWarning: is_sparse is deprecated and will be removed in a future version. Check isinstance(dtype, pd.SparseDtype) instead.
if is_sparse(pd_dtype):
/workspace/scikit-learn/sklearn/utils/validation.py:614: DeprecationWarning: is_sparse is deprecated and will be removed in a future version. Check isinstance(dtype, pd.SparseDtype) instead.
if is_sparse(pd_dtype) or not is_extension_array_dtype(pd_dtype):
/workspace/scikit-learn/sklearn/base.py:451: UserWarning: X does not have valid feature names, but IsolationForest was fitted with feature names
warnings.warn(
```

Summary

• delegate IsolationForest.score_samples validation to a new _score_samples_no_validation helper
• use the helper during fit to compute offset_ without re-validating the already-sanitized training data
• add pandas DataFrame regression tests covering fit, score, predict, and contamination='auto'

Tests

• pytest -q sklearn/ensemble/tests/test_iforest.py -k "fit_dataframe_contamination_no_warning or dataframe_then_ndarray_warns_on_score_and_predict or fit_dataframe_auto_no_warning_and_offset"
• pytest -q sklearn/ensemble/tests/test_iforest.py -k "chunks_works1 or chunks_works2"
• flake8 sklearn/ensemble/_iforest.py sklearn/ensemble/tests/test_iforest.py

Fixes #51

[casey-brooks]

Local Testing

• pytest -q sklearn/ensemble/tests/test_iforest.py -k "fit_dataframe_contamination_no_warning or dataframe_then_ndarray_warns_on_score_and_predict or fit_dataframe_auto_no_warning_and_offset" → 3 passed, 21 deselected
• pytest -q sklearn/ensemble/tests/test_iforest.py -k "chunks_works1 or chunks_works2" → 4 passed, 20 deselected
• flake8 sklearn/ensemble/_iforest.py sklearn/ensemble/tests/test_iforest.py → no issues

[noa-lucent]

The internal helper cleanly separates validation from scoring so fit no longer retriggers the feature-name warning, and the new pandas regression tests cover the relevant pathways. Looks ready to go.