fix: propagate error from Quartile for inputs too small for quartiles

Quartile computed the median of empty slices for inputs with fewer than
three elements and discarded the resulting error, returning NaN quartiles
with a nil error. The silent NaN then leaked into InterQuartileRange,
Midhinge and Trimean. Propagate the error instead, mirroring the existing
EmptyInputErr handling.

quartile.go

@@ -32,10 +32,20 @@ func Quartile(input Float64Data) (Quartiles, error) {
 		c2 = c1 + 1
 	}
 
-	// Find the Medians with the cutoff points
-	Q1, _ := Median(copy[:c1])
-	Q2, _ := Median(copy)
-	Q3, _ := Median(copy[c2:])
+	// Empty halves make Median return ErrEmptyInput; propagate it instead of
+	// returning NaN quartiles.
+	Q1, err := Median(copy[:c1])
+	if err != nil {
+		return Quartiles{}, err
+	}
+	Q2, err := Median(copy)
+	if err != nil {
+		return Quartiles{}, err
+	}
+	Q3, err := Median(copy[c2:])
+	if err != nil {
+		return Quartiles{}, err
+	}
 
 	return Quartiles{Q1, Q2, Q3}, nil
 
@@ -46,7 +56,10 @@ func InterQuartileRange(input Float64Data) (float64, error) {
 	if input.Len() == 0 {
 		return math.NaN(), EmptyInputErr
 	}
-	qs, _ := Quartile(input)
+	qs, err := Quartile(input)
+	if err != nil {
+		return math.NaN(), err
+	}
 	iqr := qs.Q3 - qs.Q1
 	return iqr, nil
 }
@@ -56,7 +69,10 @@ func Midhinge(input Float64Data) (float64, error) {
 	if input.Len() == 0 {
 		return math.NaN(), EmptyInputErr
 	}
-	qs, _ := Quartile(input)
+	qs, err := Quartile(input)
+	if err != nil {
+		return math.NaN(), err
+	}
 	mh := (qs.Q1 + qs.Q3) / 2
 	return mh, nil
 }
@@ -68,7 +84,10 @@ func Trimean(input Float64Data) (float64, error) {
 	}
 
 	c := sortedCopy(input)
-	q, _ := Quartile(c)
+	q, err := Quartile(c)
+	if err != nil {
+		return math.NaN(), err
+	}
 
 	return (q.Q1 + (q.Q2 * 2) + q.Q3) / 4, nil
 }