* Switch to marsey.cat for Snappy /u/.
camas is down, replacing it with search.marsey.cat.
Note that when looking for existing Snappy comments to test against,
it appears that something else with Snappy generation is broken.
Ex: /post/66263/-/1876803 puts an entire post URL in the author field.
This commit makes no attempt to fix this. TODO for later.
* Fix Snappy body /u/ extracting author from post URL.
Following up on 1137996f0fe7:
Issue was that author was being extracted from post.url, not href.
Given that the relevant code section is specifically for /u/s in the
body text of the submission, this was a problem.
Made the logic of the recent admig upload thread fixes (arguably)
more Pythonic, or at least less verbose.
Also, the banners path was replaced with a duplicate of the sidebars
path during the copypasta. This has been remedied.
Chose to split text into lines because Python regex supports neither
variable-width nor infinite-width negative lookbehinds. This precludes
a simple pure regex solution. Since all replacements are done at the
word level, this has no obvious ill effects.