Scope 16 / Phase 1: Compute outlier density signal in NumericProfiler

[DEVunderdog]

Parent

#105

What to build

Compute outlier_density and set NumericFlag.HighOutlierDensity in NumericProfiler. Outlier density is the fraction of non-null values that lie beyond outlier_sigma_threshold standard deviations from the column mean.

Computation (_numeric_profiler.py)

After mean and std are computed for a column:

outlier_density = count(|value − mean| > outlier_sigma_threshold × std) / n_non_null

• Store as NumericStats.outlier_density
• When outlier_density > high_outlier_density_threshold, append NumericFlag.HighOutlierDensity to NumericStats.flags
• Both thresholds are read from NumericProfileConfig (outlier_sigma_threshold, high_outlier_density_threshold)
• Skip (leave outlier_density = None, no flag) when std is None or zero (constant column)

Acceptance criteria

• A column where more than 5% of values lie beyond 3σ from the mean sets NumericFlag.HighOutlierDensity and stores the correct fraction in NumericStats.outlier_density
• A column at or below the threshold does not set the flag; outlier_density is still populated
• outlier_sigma_threshold and high_outlier_density_threshold are read from NumericProfileConfig, not hard-coded
• A constant column (zero std) skips the computation without raising — outlier_density = None, no flag
• HighOutlierDensity fires independently of KurtosisTag — a Mesokurtic column with high outlier fraction correctly receives the flag
• All existing NumericProfiler tests pass (no regressions)

Blocked by

• Blocked by Scope 16 / Phase 1: Data model — BimodalStats, TailAsymmetryTag, new NumericFlag members, NumericStats fields, NumericProfileConfig thresholds #229 (Scope 16 Phase 1 data model — NumericFlag.HighOutlierDensity, NumericStats.outlier_density must exist)