Follow-ups from the html-index source type (#137) #151

Four things came out of reviewing #137 that weren't worth holding the merge for. Listed roughly by how much they matter.

1. systemd-guides will never update. It's declared unversioned, which normally means "track tip", but it pins ref: v258. getHeadCommit on a tag returns the same commit forever, so the skip-if-unchanged check fires on every run after the first publish and latest stays v258 permanently. Either drop the ref: so it tracks tip, or keep it and add a comment saying the pin is deliberate — right now it reads like an oversight either way.

2. isZipVersionEntry is dead and changed meaning. It's no longer called outside the index.ts barrel and one test assertion, and its predicate went from "versions" in entry to source.type === "zip". Anything still importing it gets different behaviour than before. Should just be deleted.

3. registry/README.md contradicts itself. "Where the file goes" still says a non-package-manager project gets "a directory named after the project containing a single self-named file", but registry/systemd/ now holds two. The new html-index section explains the arrangement; the older rule above it doesn't know about it yet.

4. barePre loses the language fence on whitespace. The rule filters on node.firstChild?.nodeName !== "CODE", so a <pre> with any whitespace before <code class="language-x"> has a text node first and takes the bare path. The existing test only covers the no-whitespace form.

Also worth noting, not a defect: the 1,229-page crawl of freedesktop.org relies on a reviewer manually checking robots.txt, per the README. Fine at this scale, but it's prose where everything else here is enforced.