feat(html): offer an archive's entries as files, not as one page of base64

[andiwand]

🤖 Generated with Claude Code

Every entry of an archive was base64'd into the listing as a data: URL, whole,
whatever the config said — about.odt came to a 540 KB page for 172 KB of
archive. And data: cannot be navigated to at the top level in any current
browser, so the only thing you could do with an entry was save it.

Entries are HtmlResources now, like an image in a document: the locator
decides where they go, bring_offline writes them out, and the http server
serves them by path off a warmed-up render. The listing drops to 5 KB beside the
files it names, and each entry is a real URL.

In the listing, the path is the link that opens the file and the glyph
beside it saves it — one glyph, since a second would have had to mean "open" and
be told apart from the first. An embedded entry has no URL to open, so there the
path is plain text.

Directories are gone. Every entry names its whole path, so a row for the
directory above it repeated what the rows below already said, and it was the one
row with nothing to open, save or measure.

Untrusted names stay inline. The archive names the entry, so neither its
name nor its path is trusted with a location: a .. that would climb out of the
output directory, or a files.html that would overwrite the listing, gets none.

HtmlResourceType gains file, appended — the bindings mirror it by ordinal.

Testing

Full suite: 818 passed, 8 skipped, no failures. html.archive_listing now
asserts no data: in the page, a link per entry, and the entry written beside
the listing. Reference output regenerated and pinned: zip/small.zip gains
hello.txt and nested/note.txt next to a listing that shrank from 889 to
864 bytes.

Rebase note

The branch was stacked on #670's pre-merge commits. It is rebased onto main
with only its own commit; the archive listing's service picked up the
resource-serving fix that landed with #670, so warmup() now covers both the
stylesheet and the entries and locate_filesystem_resources is gone.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7c369c6ead

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".