extractxml: fix missing world.html data

Python requires an explicit close() call to ensure that all data passed by write() before is also ending up in the filesystem.

Separate commit also fix output to stdout, as interesting for testing.

Beware: I do not really write Python code myself, did those fixes mainly by try&error and search hits ;)

Not sure if this needs to be backported, people seem to not have missed the missing translations for the world description so far.