The api data is unavailable when we can't scrape the uuid, either
because the work is deleted or because the work is actually a sta.sh
upload and we weren't given the sta.sh page in the referer url.
Make <role>_only methods check the role first and ip bans last. This
avoids hitting the database for anonymous users, since they'll always
fail the is_<role>? check before the ip check.
* Replace AnonymousUser null object with a readonly, unpersisted User object.
* Default always_resize_images to true (previously it was true for
anonymous users, but false for new members).
* Default comment_threshold to -1 for anonymous users (previously it was
0 for anonymous but -1 for new members).
Instead of calling /sources.json and rendering the "Fetch source data"
box client-side in Javascript, call /sources.js so we can render the
html server-side.
* `legacy_normalize` came from c6012535, which is no longer a problem.
* `normalize_for_search` is only used for "[mass edit]" links
in artist entries. These links are a shortcut for performing a
`-artist_name source:<artist_url> -> artist_name` mass edit to tag
untagged artists, but this won't work for most sites these days.
Captioned post previews (previews with the pool name, similarity, or
size beneath) need to have `height: auto` set, otherwise they'll default
to `height: 154px` (just enough for the image) and the caption won't be visible.
* On the /uploads/new page, instead of just showing a "This post has
probably already been uploaded" message, show the actual thumbnails of
posts having the same source as what the user is trying to upload.
* Move the iqdb results section up top, beside the related posts section.
Also makes it so that using the bookmarklet always triggers async upload
preprocessing. Before it was only triggered when the source passed a
dupe check, but that check was inaccurate (#3873).
Fixes the '.current' class not being set on the 'Forum' link when
viewing the /forum_topics page. As a result, the link wasn't bolded and
didn't have the background set.
