'latin-1' and 'utf-8'.

These are optimized in the Python Unicode implementation
to result in more direct processing, bypassing the codec
registry.

Also see issue11303.