Fix the benchmark when we have a real registered factory

Any slowdown appears to be small so far.

+------------------------------+----------------+-----------------------------+
| Benchmark                    | 27_master_list | 27_issue51_1                |
+==============================+================+=============================+
| __main__: toExternalObject   | 405 us         | 437 us: 1.08x slower (+8%)  |
+------------------------------+----------------+-----------------------------+
| __main__: fromExternalObject | 1.37 ms        | 1.41 ms: 1.03x slower (+3%) |
+------------------------------+----------------+-----------------------------+

src/nti/externalization/datastructures.py

@@ -443,9 +443,10 @@ def get_object_to_update(self, key, value, registry):
                 field = self._iface[key]
                 # XXX: Maybe this should be a string naming a factory we find in the registry?
                 # Or a string giving the dottedname to a class we resolve at runtime (and reify)
-                return field.getTaggedValue('__external_factory__')
+                factory = field.getTaggedValue('__external_factory__')
             except KeyError:
-                return None
+                pass
+        return factory
 
     def updateFromExternalObject(self, parsed, *unused_args, **unused_kwargs):
         result = AbstractDynamicObjectIO._updateFromExternalObject(self, parsed)

src/nti/externalization/tests/benchmarks/bm_simple_iface_list.py

@@ -40,7 +40,6 @@ def main(runner=None):
     obj.the_objects = [DerivedWithOneTextField(text=u"This is some text " + str(i))
                        for i in range(10)]
 
-
     if '--profile' in sys.argv:
         profile(100, obj)
         return