The test checks this by verifying that the newlines in the "pre" post are converted to spaces.

This allows posts with autogenerated excerpts (that may have images or HTML in them) to keep that HTML intact in the summary portion of the Atom feed. The default remains as having the HTML be stripped and the whitespace normalized.

This has some benefits, because the default value for a post excerpt is for it to be the entire first paragraph of each post. It's not unheard of for it to have plenty of HTML formatting and other tags.

In [my site][1], for example, post excerpts often have images, and I'd very much prefer keeping those images intact in the feed, even if using `excerpt_only` feed generation.

Note that this passes the W3C validator with no warnings, since feed.xml already uses `type="html"` in its `<summary>` tags.

[1]: https://www.kronopath.com/blog/

This naming is more consistent with the "excerpts_only" flag, though it may be slightly more confusing given that this also applies to the post's "description" option. Still, consistency is key.

Apparently some of the test cases are failing because instead of the "pre" post being tested, it's instead picking up the "liquid" post from just before it.

I suspect that this might be happening because the test machines are maybe configured with a date and time before 2016-04-25, meaning that they don't pick up the "author-reference" post, leading to an off-by-one error. That's my best guess, anyway.

It's not the most robust answer, since it would also likely mean that in these machines the "puts the latest 2 the posts in the feed.xml file" test would also fail. So maybe I'm wrong.

Still, to try to fix this, instead of counting the posts from the most recent, the test now instead counts from the oldest, which should hopefully be more robust to this issue.

Apparently the indices just aren't consistent on jekyll-feed's test machines. Neither `feed.items[3]` nor `feed.items[-7]` can be reliably said to be the "pre" post which I use for testing in this PR.

Instead, let's just search through all the posts for whichever one is titled "Pre" and test against that one.