change test to use html normalizer when comparing html output. Update to Markdown2 parser changed text output keeping same html semantics. Broke test_string_markdown_code_block_attribute test. I hand patched it to get tests working but it needed a better solution. Write a simple html normalizer using HTMLParser so I don't need third party (lxml, beautifulsoup) library to clean up the test. Use the normalizer to parser the expected result and the result returned by the various markdown libraries. Hopefully this will make the test less fragile. This can have multiple uses in template testing where html is compared. I expect to have to change html_norm.py to make test writing easier in the future.