Fix RFI bug: Handle scalar/array diffs without to_numpy() method (AttributeError: 'float' object has no attribute 'to_numpy')

[jemus42]

• Use np.asarray() for scalars and numpy arrays
• Use to_numpy() only for pandas objects that have it

Disclaimer: I barely know what I'm doing but doing this... helped?

Reprex:

import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error
from fippy.explainers import Explainer
from fippy.samplers import GaussianSampler

# Generate simple test data
np.random.seed(123)
n_samples = 100
X = pd.DataFrame({
    'feature1': np.random.normal(0, 1, n_samples),
    'feature2': np.random.normal(0, 1, n_samples),
    'feature3': np.random.normal(0, 1, n_samples),
})
# Simple linear relationship
y = pd.Series(2 * X['feature1'] + X['feature2'] + np.random.normal(0, 0.1, n_samples))

# Train/test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=123)

# Train model
model = RandomForestRegressor(n_estimators=50, random_state=123)
model.fit(X_train, y_train)

# Set up fippy explainer
sampler = GaussianSampler(X_train)
explainer = Explainer(model.predict, X_train, loss=mean_squared_error, sampler=sampler)

# RFI with conditioning set - this will trigger the bug
conditioning_set = ["feature1"]  # Condition on feature1

explainer.rfi(conditioning_set, X_test.head(10), y_test.head(10))