closes #18

if the photo has a location, the timezone is set on the published date

If the input URL contains a fragment, finds the DOM tree at that ID and runs the subtree through the mf2 parser.

closes #15

* fields that should be URLs will now be omitted if the value was not a URL, such as when the value is `javascript:alert()`
* makes Mf2 class slightly more self-contained by duplicating the URL helper functions into it
* fixes tests to not cache responses in memcache

closes #9

see http://microformats.org/wiki/microformats2-parsing-issues#ignore_u-camelCase_properties for background

allow only a basic set of tags, and remove any non-mf2 classes

closes #2

sanitize the HTML returned in the content property. allows a common set of HTML tags.

for #2

if a property such as `in-reply-to` is an h-cite, the URL is still returned as the `in-reply-to` value, and the h-cite object is available in a different part of the response.

closes #6

file_get_contents is dumb. I hope this isn't a permanent solution.

see https://cloud.google.com/appengine/docs/php/urlfetch/

* extracts mf2 post contents from pages
* implements authorship discovery to find author info for the URL