[gatsby-remark-copy-linked-files] Fix regression from #2871 with RegExp cannons

[fk]

Sooooo…I broke resource references being updated for HTML-in-Markdown links, videos, and images in #2871 when trying to fix #2696 by removing writing back the HTML node.value via Cheerio’s .html(). 🙈

Cheerio closes open HTML tags, which in our case caused the problems described in #2696—an HTML node a with value <a href=…> would be written back as <a href=…></a>, putting the link children after the link:

const $ = cheerio.load('<a href="http://example.com/">')
$.html()
//=> <a href="http://example.com/"></a>

Not serializing the Cheerio object of our node back to node.value of course resulted in the resource attributes altered with it being lost.

---

Since I couldn’t figure out how to use Cheerio nor Rehype/unified to a) comfortably and relatively safely update the src and href attributes while b) returning just a fragment of the HTML node (<a href=...> instead of <a href=...></a>) , here’s an ugly beginner fix 😬 that replaces the resource links directly in node.value…

All in all this is merely trying to follow up on the mess I made in #2871 as good as my available time currently allows – not convinced about this approach at all, so I left the replace stuff as dumb as possible for now and added a new example instead: examples/using-remark-copy-linked-files that hopefully makes it a little easier for other people to take a look. ⛑ 🙏

Here's the new example site's main Markdown file, exploring the AST might help understanding my blatherings: https://astexplorer.net/#/gist/307f2f3226954fa34304503a8d9c6cde/c52af3223c9e2b3a14f21ec0df00f143cd688037

[gatsbybot]

Deploy preview ready!

Built with commit 78f43d7

https://deploy-preview-2901--gatsbygram.netlify.com

[gatsbybot]

Deploy preview ready!

Built with commit 78f43d7

https://deploy-preview-2901--using-drupal.netlify.com

[KyleAMathews]

This is great! Cheerio is nice but yeah, didn't turn out to be a great fit for us as they want to create "well-formed" HTML which doesn't always jive with user input or perhaps how Remark parses things. So yeah, Regex seems like the way to go to ensure we're not breaking people's HTML.