docs(adr): shared spaces, a published studyset and the copy

ADR 0044, D27 and D28.

- D27: sharing a studyset with a publication anywhere in its tree is
  refused for now. Publishing reads a release from the person's
  readable rows on the instance, and a shared studyset is sealed in
  another space, so it could take no release; a later step has the
  device hand over a release's content and re-points the publication.
- D28 (the agent's, to be confirmed): the copy is the archive's export
  and import, so progress, tags, stars and the archive follow, each
  item keeping its link to Explore and each card its origin.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>

docs/adr/0044-shared-spaces-and-collaboration.md

@@ -3,7 +3,8 @@
 - **Status:** decided 10 Nov 2026 by the developer: D8 is their own request, D9 and D10
   their choices over the agent's recommendations (all ticked at first; agents editing
   directly), D1 to D7 the agent's recommendations, which they chose; D11 to D13 the
-  agent's recommendations, which they chose on 11 Nov 2026, D14 too, and D15 to D26.
+  agent's recommendations, which they chose on 11 Nov 2026, D14 too, and D15 to D26;
+  D27 their choice; D28 the agent's, to be confirmed.
   Steps 1 to 3 built; the rest below.
 - **Builds on:** ADR 0005 (publishing, reports and suggestions), ADR 0014 (the module),
   ADR 0016 (one account across an instance), ADR 0017 (key pairs), ADR 0019 (private
@@ -143,6 +144,23 @@ member reads everything in the space.
   shared space would show members each other's reviews; teaching the engine to link
   across spaces goes against its design.
 
+- **D27.** **A published studyset isn't shared, for now** (decided 11 Nov 2026, the
+  developer's choice of the agent's corrected recommendation): publishing reads a
+  release from the person's readable rows on the instance (`readReleaseFiles`,
+  refusing a private studyset), and a shared studyset is sealed (D21) in another
+  space, so it could take no release. Sharing a studyset with a publication anywhere in
+  its tree, withdrawn or not, is refused, saying why, until publishing takes a
+  release's content from the device, a step of its own that re-points the publication
+  to the shared studyset then.
+- **D28.** **The copy is the archive's export and import** (11 Nov 2026, the agent's
+  choice, for the developer to confirm): the tree written out with its progress, as a
+  backup is, and read back under new IDs into the space through the person's
+  repositories, so reviews, answers, readings, tags by name, stars and the archive
+  follow as an import's do and no rewrite of its own is needed; unlike a backup, it
+  keeps each item's link to Explore (`published`) and each card's origin, as the
+  studyset is the same one moved. The old tree is deleted for good once the copy is
+  whole; a copy failing or aborted deletes what it wrote and the space it made.
+
 ## Steps
 
 1. **Share with the community's confirmation** (D8, D9), needing nothing else here.