test: establish better baseline behavior for MS Word's html format

because this is about to change in libxml 2.9.12

test/xml/test_node.rb

@@ -1049,18 +1049,22 @@ def test_namespace_without_an_href_on_html_node
           #  describe how we handle microsoft word's HTML formatting.
           #  this test is descriptive, not prescriptive.
           #
-          skip_unless_libxml2("Xerces handles this edge case completely differently")
-
-          xml = Nokogiri::HTML.parse(<<~EOF)
+          html = Nokogiri::HTML.parse(<<~EOF)
             <div><o:p>foo</o:p></div>
           EOF
-
-          node = xml.at("p")
+          node = html.at("div").children.first
           assert_not_nil(node)
 
-          assert_equal(1, node.namespaces.keys.size)
-          assert(node.namespaces.has_key?('xmlns:o'))
-          assert_nil(node.namespaces['xmlns:o'])
+          if Nokogiri.uses_libxml?
+            assert_equal(1, node.namespaces.keys.size)
+            assert(node.namespaces.has_key?('xmlns:o'))
+            assert_nil(node.namespaces['xmlns:o'])
+            assert_equal("<p>foo</p>", node.to_html)
+          else
+            # Xerces does something completely different
+            assert_empty(node.namespaces.keys)
+            assert_equal("<o:p>foo</o:p>", node.to_html)
+          end
         end
 
         def test_xpath_results_have_document_and_are_decorated