Implement an export subcommand (page + attachments to the filesystem)

[willkg]

Add an export subcommand that writes a Confluence page and its attachments to a directory, mirroring pchuri/confluence-cli's export. This builds on read (page → markdown/storage) by also downloading attachments and laying everything out on disk so the exported markdown's image links resolve locally.

Command

markfluence export ARG [flags]

ARG is a numeric page id or a Confluence page URL (same resolver as read).

Flags (mirroring confluence-cli)

• --dest <dir> — output directory (default .).
• --file <name> — content filename (default derived from title/id, e.g. <slug>.md).
• --attachments-dir <name> — attachments subdirectory (default e.g. attachments/).
• --format markdown|storage — content format (default markdown), reusing read's conversion. (confluence-cli also has html/text; markfluence has neither yet — out of scope unless we add them.)
• --pattern <glob> — only export attachments matching the glob.
• --referenced-only — only download attachments actually referenced in the body.
• --skip-attachments — write the content file only.

Filesystem output

<dest>/
  <file>.md                # page content (markdown by default), with frontmatter
  <attachments-dir>/       # downloaded attachment files
    ...

New capability: downloading attachments

markfluence today lists (ListAttachments) and uploads attachments but never downloads their bytes. Export needs a client method to fetch an attachment's binary (v1 .../download or the attachment _links.download), plus listing/pagination.

The image-link reconciliation (the crux)

read's markdown emits the flattened attachment filename (assets/x.png → assets_x.png) as the image src. For an export to be self-contained, the downloaded files and the markdown srcs must line up — e.g. download each attachment into <attachments-dir>/ and rewrite image srcs to <attachments-dir>/<filename>. This also closes the local-image round-trip gap noted during the read work (reading a page couldn't recover image bytes/paths).

Open questions

• Default content filename: slugified title, page id, or require --file?
• Attachment name collisions (Confluence allows multiple attachments; flattened names could clash) — how to disambiguate.
• --referenced-only scope: only ac:image/ri:attachment references, or also links to attachments?
• Should export also emit enough for a later create/update (frontmatter with page_id), so an exported tree round-trips back? (It would, via the markdown frontmatter.)
• Multi-page export (a page + its descendants) — out of scope for v1, or a --recursive/depth flag later?

References

• confluence-cli export: https://github.com/pchuri/confluence-cli#export-a-page-with-attachments
• Builds on the read subcommand and the storage→markdown converter (merged).