Fixes #4791
Previously, our rich text conversion functions handled the case where a document link specified an ID which is not found in the database. However, they failed with a KeyError when the id attribute was missing completely; links of this second type would occur whenever a link of the first type was re-saved from the Draftail editor. The fix is two-fold:
1) Catch the "missing ID attribute" case - in this case, the resulting link will be missing both the href and id attributes
2) Update the handling of the "ID present but document not found" case so that the id attribute survives the round-trip to the editor and back. The final link as rendered on the front-end will still be an attribute-less <a> element, but the id will be retained in the database (and in the versions rendered within rich text editors) which may be useful for troubleshooting.
Currently queries executed in the hooks don't run in the transaction
with the page deletion query and it's harder to write hook without
copying the whole view if you want to keep queries running in the hooks
integral with page deletion.
According to a google search I just did, it seems a lot of people have forgotten to add ``Disallow: /admin`` in their robots.txt (or forgot to add robots.txt) at all.
Adding this meta tag into the head of all admin pages should prevent any admin pages being indexed even if this was missed.
