@spec evaluate(Aludel.Evals.Metric.Context.t() | String.t(), map()) :: map()

Evaluates an assertion against generated output or a complete metric context.

Returns the legacy string-keyed result map used by suite runs and exports. An unknown metric type produces a failed result instead of raising.

@spec score_for_results([map()]) :: float() | nil

Calculates the mean numeric score from assertion result maps.

Returns nil when the list is empty or contains no numeric "score" values.