Fix for #7369: return scalar for sampler stats

Author: asifzubair

pymc/step_methods/metropolis.py

@@ -1091,7 +1091,7 @@ class DEMetropolisZ(ArrayStepShared):
         "accept": (np.float64, []),
         "accepted": (bool, []),
         "tune": (bool, []),
-        "scaling": (np.float64, []),
+        "scaling": (np.float64, [None]),
         "lambda": (np.float64, []),
     }
 
@@ -1214,7 +1214,7 @@ def astep(self, q0: RaveledVars) -> tuple[RaveledVars, StatsType]:
 
         stats = {
             "tune": self.tune,
-            "scaling": self.scaling,
+            "scaling": np.mean(self.scaling),
             "lambda": self.lamb,
             "accept": np.exp(accept),
             "accepted": accepted,