Do not consider namespaces missing if they are already emitted in kee…

[thetrime]

…p_prefix mode

Without this, you get

?- xml_write(user_error, element(ns(foo, 'http://www.example.com'):bar, [], [element(ns(foo, 'http://www.example.com'):baz, [], [])]), []).
<?xml version="1.0" encoding="UTF-8"?>

<foo:bar
    xmlns:xns22="http://www.example.com"
    xmlns:xns21="http://www.example.com">
  <foo:baz/>
</foo:bar>
true.

With the patch, you get

<foo:bar xmlns:xns19="http://www.google.com">
  <foo:baz/>
</foo:bar>

I think this might still be wrong, but it's not as wrong?

[thetrime]

Actually it's plenty wrong. The right answer, is, of course

<foo:bar xmlns:foo="http://www.google.com">
  <foo:baz/>
</foo:bar>

I'll keep digging here

[thetrime]

Ok, I think this looks right now. At least, it works for my test case, and it still generates the right thing when removing the ns/2 from the test case.

[JanWielemaker]

Thanks. Merged after squash and edit of commit message.