don't calculate std_dev with kwargs inputs, cleanup f_with_affine_out…

…put a little

uncertainties/ops.py

@@ -527,29 +527,25 @@ def f_with_affine_output(*args, **kwargs):
 
         for pos in pos_w_uncert:
             arg_uncertainty = args[pos].uncertainty
-            if arg_uncertainty.ucombo_tuple:
-                uncertainty += (
-                    derivatives_args_index[pos](*args_values, **kwargs)
-                    * arg_uncertainty
-                )
+            if arg_uncertainty:
+                derivative_val = derivatives_args_index[pos](*args_values, **kwargs)
+                uncertainty += derivative_val * arg_uncertainty
 
         for name in names_w_uncert:
             # Optimization: caching of the automatic numerical
             # derivatives for keyword arguments that are
             # discovered. This gives a speedup when the original
             # function is called repeatedly with the same keyword
             # arguments:
-            if kwargs_uncert_values[name].s == 0:
+            if not kwargs_uncert_values[name].uncertainty:
                 continue
-            derivative = derivatives_all_kwargs.setdefault(
+            derivative_func = derivatives_all_kwargs.setdefault(
                 name,
                 # Derivative never needed before:
                 partial_derivative(f, name),
             )
-            uncertainty += (
-                derivative(*args_values, **kwargs)
-                * kwargs_uncert_values[name].uncertainty
-            )
+            derivative_val = derivative_func(*args_values, **kwargs)
+            uncertainty += derivative_val * kwargs_uncert_values[name].uncertainty
 
         # The function now returns the necessary linear approximation
         # to the function: