This gets rid of the use of javascript for converting node/post description.
Now we only use markdown.js for real time as-we-type stuff, like node/post
editing or commenting.
Now we can access pages with the following url
/p/<project_url>/<page-url>. Internally we use the existing view_node,
but if we detect that the node_id is not an object id we try to treat
it as a page url and therefore we try to define node and project using
render_node_page().
