Fix the problem where Cohen's h was not calculated correctly

for the case of nonbinary samples.
The bug was fixed by using changing the return value.

Author: Jacobluke-

dabest/_stats_tools/effsize.py

@@ -259,7 +259,7 @@ def cohens_h(control, test):
     try:
         pd.unique(control)==np.array([0,1]).all()==False and (pd.unique(test)==np.array([0,1])).all()==False
     except:
-        pass
+        return None
 
     # Convert to numpy arrays for speed.
     # NaNs are automatically dropped.