Fix #334.

This fixes #334 by sorting everything before writing it out to the
localization file. Note that we need to sort each step along the way
because the nested dicts are not sorted by a call to `sorted()` - it
appears to only sort the key.

korman/exporter/locman.py

@@ -150,7 +150,7 @@ class LocalizationConverter:
 
         def iter_element(element):
             if language_name is None:
-                yield from element.items()
+                yield from sorted(element.items())
             else:
                 yield language_name, element
 
@@ -160,9 +160,9 @@ class LocalizationConverter:
             write_line("<localizations>")
             write_line("<age name=\"{}\">", self._age_name, indent=1)
 
-            for set_name, elements in sets.items():
+            for set_name, elements in sorted(sets.items()):
                 write_line("<set name=\"{}\">", set_name, indent=2)
-                for element_name, value in elements.items():
+                for element_name, value in sorted(elements.items()):
                     write_line("<element name=\"{}\">", element_name, indent=3)
                     for translation_language, translation_value in iter_element(value):
                         if _ESHTML_REGEX.search(translation_value):