feat(discussions): questions on published studysets, answered by learners

A learner asks about a published studyset, a topic in it, or one card,
note or step of a release, and others answer, as on Stack Exchange
(ADR 0013): questions and answers in the app's editor without images,
sounds or videos, comments of up to 600 characters under the question
and each answer. There are no votes: the asker accepts an answer, which
comes first, then the oldest; the author's posts are marked. The
instance reads from git's file list which topics a question's target
sits in, so a topic lists what is asked within it and below it.

Posts live in two server-only tables, written through endpoints, so the
synced tables stay as they are and v0.17.0 keeps syncing. Everyone
signed in reads; writing needs an identity and a public profile; the
studyset's author hides posts and blocks people with feedback's blocks;
a withdrawn studyset takes no new questions; anyone writes at most 30
posts an hour. A post with others' replies under it is emptied when
deleted, else deleted; deletePerson deletes what a person wrote the same
way, and a published studyset's deletion takes its discussions with it.

The migration adds Serverpod's gen_random_uuid_v7(), discussion_post
and discussion_seen with their indexes and cascading foreign keys, and
changes nothing else.

In the app, discussions open from a copy's studyset or topic menu, a
studied card's "Ask about this", Explore's summary and the Publish
sheet; the Profile tab's inbox lists those the person follows, and its
count includes what is new in them.

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/0013-discussions.md

@@ -0,0 +1,78 @@
+# ADR 0013: Discussions on published studysets
+
+- **Status:** accepted
+- **Date:** 2026-10-15
+
+## Context
+
+A published studyset has reports and suggestions (ADR 0005, rule 11): a line between one
+writer and its author, written as rows in each writer's own synced space and gathered by
+the instance as they are read. Learners also have questions about what they learn, and
+other learners can answer them, as on Stack Exchange; the developer's example was
+"Meaning of 青い衝撃 in this situation?" in Japanese › Grammar. The working notes (11 Oct)
+planned these as discussions "built on the feedback threads", the step before federation.
+
+Feedback's storage doesn't fit them well: many people answer, the asker accepts one,
+posts are edited, and lists are ordered and paged by activity. Rows spread across
+writers' spaces could only encode acceptance in decision rows, and could never be edited
+(synced content is only ever blanked). New synced tables would make every app older than
+the change stop syncing (the sync engine refuses an app whose synced tables differ).
+
+## Decision
+
+The developer decided (2026-10-15), each as recommended:
+
+1. **Server tables, not synced.** `discussion_post` holds questions, answers and
+   comments, `discussion_seen` when each person last read a discussion; both are
+   server-only, written through endpoints, online only. Counting, ordering, accepting and
+   editing are plain queries, posts are ready to federate later (ADR 0005, stages 5 and
+   6), and the synced tables stay as they are, so older apps keep syncing.
+2. **Questions, answers and comments**, as Stack Exchange has them: comments go under the
+   question and under each answer, never under a comment; questions and answers are rich
+   text in the app's editor (without images, sounds or videos: the person's files are
+   theirs alone, and nobody else could load them), comments plain text of up to 600
+   characters.
+3. **No votes.** The accepted answer first, then the oldest first; the author's posts
+   marked. Votes cost nothing to fake (rule 8 of ADR 0005), so they count for nothing.
+   The asker accepts an answer, and can take it back.
+4. **About the studyset, a topic or an item.** A question names what it is about as the
+   release names it: the published studyset or a topic in it, a card, a note or a step,
+   by its id in a release. The instance reads where that sits in the release (git's file
+   list alone) and keeps the topics it is within, so a topic lists what is asked within
+   it and in the topics below it. A copy's items know their ids and releases, so the app
+   asks about them directly; Explore asks about the studyset as a whole, naming no item.
+5. **Deleting keeps others' replies.** A post nobody else replied under goes for good; a
+   post with others' replies under it stays for them, its title, text and writer gone
+   ("Deleted question"). Deleting a person deletes what they wrote the same way, in
+   `deletePerson`'s transaction; deleting a published studyset deletes its discussions.
+6. **The feedback rules stay**: everyone signed in reads; writing needs an identity and a
+   public profile (posts name their writer); the studyset's author hides posts (then
+   only their writer and the author see them) and blocks people (the same blocks as
+   feedback's: what a blocked person writes shows to them alone). A withdrawn studyset
+   keeps its discussions readable and answerable, but takes no new questions. A person
+   writes at most 30 posts an hour.
+7. **Where it shows**: a copy's menu opens the discussions within that studyset or topic;
+   a studied card from a release offers "Ask about this"; Explore's summary and the
+   author's Publish sheet open the whole studyset's; the person's inbox (Profile) lists
+   the discussions they follow (on their studysets, or asked, answered or commented in),
+   what others wrote since they last read one marked and counted on the Profile tab.
+
+## Consequences
+
+- The synced tables don't change: v0.17.0 keeps syncing with an instance that has
+  discussions.
+- Writing needs the instance; a question being written stays in the composer meanwhile.
+- Moderation is the author's alone; an instance's operator has no tools for it yet.
+- Federation (Fedify, ADR 0005 rule 7) can announce posts as the instance holds them.
+
+## Alternatives considered
+
+- **Rows in each writer's synced space, as feedback**: offline writing, one pattern for
+  both; but no edits, and acceptance and ordering pieced together from many spaces.
+- **New synced tables**: offline and clean, but older apps stop syncing until updated.
+- **Votes ordering answers**, or shown without ordering: the familiar way, but votes are
+  free to fake, and Explore already refuses to rank by them.
+- **Flat replies** (a forum thread): simpler, but no accepted answer, and short remarks mix
+  with answers.
+- **Everything goes with a deleted question**: simpler, but people's answers would vanish
+  with someone else's choice.