The current news-rss.xml fails http://www.feedvalidator.org for a couple
of reasons, making feed readers like Liferea misparse it. This patch
fixes one of those issues.

As @edolstra noted in #90, HTML is not allowed directly in the RSS
`<description>` tag. Instead it's supposed to be a text node, where the
text can be parsed as HTML. So e.g. an `<a>` tag needs to be written in
the feed as `&lt;a&gt;`.

This could be handled more easily in an Atom feed, which allows tags
from namespaces like XHTML to be embedded directly as long as the
namespaces are declared properly.

This could also be simpler if xsltproc supported the XPath Functions 3.0
standard, which defines a `serialize` function that does what the
mode="serialize" template does in this patch.

The current news-rss.xml fails http://www.feedvalidator.org for a couple
of reasons, making feed readers like Liferea misparse it. This patch
fixes one of those issues.

The RSS pubDate element is required to be an RFC-822 date-time. The
entries in news.xml did not conform to that specification for two
reasons: a missing comma, and swapping the month and day fields.

So I fixed both of those issues in news.xml, and then fixed news.xsl to
extract the date substring at its new offset.

This does mean that the HTML version of the news has its month and day
swapped now, and frankly I liked it better in the previous month-day
order.

If desired the previous output can be recovered, by either continuing
using the substring-pasting approach but chopping up the substrings
further, or changing news.xml to use ISO 8601 date-times and using the
http://exslt.org/date/ extension functions.

The channel's <link> tag should point to an HTML version of the same
content as the feed, so it's better to link to /news.html than to /.

In addition, the RSS specification says:

"the image <title> and <link> should have the same value as the
channel's <title> and <link>."

http://www.rssboard.org/rss-specification#ltimagegtSubelementOfLtchannelgt

Finally, feedvalidator.org says the feed document should include a link
to the canonical URL for that feed:

http://www.feedvalidator.org/docs/warning/MissingAtomSelfLink.html

The item's <link> tag should point to an HTML version of the same
content as the item. However, there is no URL for each item in the
current site, so there's no good place to link to. Fortunately, the link
tag is not required when a description is present, so we can just drop
that tag entirely.

Also, each item should have some sort of unique ID in the <guid> tag.
If a feed reader sees an item that's different from any item it has seen
before, then the guid allows the reader to distinguish between new items
and items that have been edited.

http://www.feedvalidator.org/docs/warning/MissingGuid.html

Since it has almost never happened that two items were posted with the
same pubDate, I've chosen to use the pubDate itself as the guid. There
were three duplicates, but I've made them unique by adding one second to
the pubDates of the three duplicates that appear earlier in the file, so
that sorting by descending pubDate would leave the file order unchanged.

First, I removed the maxItem limit from the RSS feed. (It's still used
for the HTML version.) The maxItem limit of 1000 was effectively
infinite at the rate news items are being published, and each item is
quite small so there isn't much bandwidth cost to publishing a lot of
them anyway.

Second, since the feed does contain every entry ever posted, it is a
valid "complete feed" in the sense of RFC5005 section 2, so add the tag
marking it as such.

If the RSS feed ever gets too large, then RFC5005 section 4 describes
how to paginate it into archived feeds.