Fix detect_outliers_ecod to handle pandas Series and missing values

[Copilot]

detect_outliers_ecod used raw integer indexing (N[0], N[i], slices) which breaks silently when passed a pandas Series with a non-default index (label-based access). Additionally, NaN inputs caused obscure ECOD fitting errors with no actionable message.

Changes

• Input coercion: Convert N to np.ndarray via np.asarray(N, dtype=float) at function entry — fixes label-vs-position indexing for any pandas Series index
• NaN guard: Raise ValueError with a clear message when input contains NaN values, rather than letting ECOD fail opaquely

import pandas as pd
import numpy as np
from growthcurves.preprocessing import detect_outliers_ecod

# Previously broken: non-default index causes KeyError or wrong results
s = pd.Series([0.1, 0.2, 0.3, 0.5, 0.8], index=range(5, 10))
mask = detect_outliers_ecod(s)  # now works correctly

# Missing values now raise a clear error
s_nan = pd.Series([0.1, np.nan, 0.3])
detect_outliers_ecod(s_nan)  # raises ValueError: "...does not support missing values (NaN)..."

Tests added

• test_ecod_pandas_series_default_index — baseline Series case
• test_ecod_pandas_series_non_default_index — previously broken path
• test_ecod_numpy_array — regression guard
• test_ecod_missing_values_raises — NaN in ndarray raises ValueError
• test_ecod_missing_values_in_series_raises — NaN in Series raises ValueError