feat(data): retain conversations on archive() + expose archived filter + seed

[ilmoniemi]

User Story

As an app developer wiring the Archived Discussions UI, I want archived conversations to remain observable through the repository (with a flag, not by deletion), so the UI layer can filter and render an "Archived discussions" list against a realistic data shape.

Context

The forthcoming Archived Discussions view (split from #77) needs to observe conversations matching isPromoted = false AND archived = true and offer a restore action. The repository surface introduced in #3 already exposes archive(conversationId), but the current FakeConversationRepository implementation removes the conversation from the in-memory store entirely (records - conversationId), so archived state is unobservable today and the existing filter enum (All, Channels, Discussions) has no way to surface it.

This ticket converts archive into flag-based retention so the UI slice has something to observe. The trivial inverse — unarchive(conversationId) — lands in a separate sibling ticket that this one unblocks.

Acceptance Criteria

• Conversation (in data/model/Conversation.kt) gains an archived: Boolean = false field
• FakeConversationRepository.archive(conversationId) retains the conversation with archived = true instead of removing it from the store
• ConversationRepository exposes an archived-only observable stream — exact shape is the architect's call (new ConversationFilter variant, additional argument, or other), but Discussions must continue to mean "non-promoted and not archived" so the existing Recent Discussions list stays correct
• FakeConversationRepository seeds at least one already-archived discussion (isPromoted = false, archived = true) so the archived stream is non-empty for manual verification
• Unit tests cover: the seeded archived discussion appears in the archived stream and not in the Discussions stream; the existing archive_removes_from_observeConversations test is rewritten to assert the new flag-based semantics (archived items leave the Discussions stream and appear in the archived stream)

Technical Notes

• The repository interface in data/repository/ConversationRepository.kt uses raw String for conversationId parameters today; there is no ConversationId value type yet. Stay consistent with the existing surface — do not introduce a value type as part of this ticket.
• Any seed-count assertions in FakeConversationRepositoryTest (e.g. assertEquals(5, ...)) need updating to reflect the new seeded archived discussion.

Size Estimate

XS

Split from

#77

[ilmoniemi]

PO rework + split. The original body claimed:

1. A ConversationId value type from feat(data): add Conversation, Session, Message data classes #2 — there is no such type; the interface uses raw String.
2. That unarchive clears an archived flag — there is no flag on Conversation; the current FakeConversationRepository.archive() actually removes the conversation from the in-memory store (records - conversationId), so archived state is unobservable.

To make unarchive even meaningful, archive's semantics have to change to flag-based retention first, plus an archived-observable stream has to be exposed, plus a seed has to land for manual verification. That's a coherent foundation slice on its own (5 ACs, XS).

Pairing the foundation with the inverse unarchive primitive would push the combined ticket past the >5 AC red line and force an "and" into the sentence — so split. This ticket (#93) is now the foundation; #96 carries the unarchive primitive and is blocked-by this one. The downstream UI ticket #94 is now blocked-by both.

[ilmoniemi]

🤖 Product Owner — creates structured issues

po agent has completed work on this ticket.

Agent output (click to expand)

Rework + split complete.

**Why this needed rework.** The original #93 body claimed two things the codebase doesn't support: a `ConversationId` value type (the interface uses raw `String`) and an `archived` flag to clear (current `FakeConversationRepository.archive()` *removes* the conversation, no flag exists). The trivial "add `unarchive()`" framing was unworkable.

**The split.**
- **#93 (refined, foundation):** Add `archived: Boolean` to `Conversation`, convert `archive()` to flag-based retention, expose an archived-observable stream (architect picks the filter shape, but `Discussions` must keep meaning "non-promoted AND not archived"), seed one already-archived discussion, update the existing `archive_removes_from_observeConversations` test to the new semantics. 5 ACs, XS. Title changed to "feat(data): retain conversations on archive() + expose archived filter + seed".
- **#96 (new sibling):** Add `suspend fun unarchive(conversationId: String)` to the interface + Fake impl that clears the flag + round-trip test. 3 ACs, XS. `blocked-by` #93. Positioned in Backlog immediately after #93.
- **#94 (downstream UI):** Now `blocked-by` both #93 and #96 (was previously only #93). Commented to explain the re-point.

Combining all three concerns into one ticket would have hit 7 ACs (>5 red line) and required "and" in the description, both of which are mechanical split triggers per the PO sizing guide.

Ready for human review. Move to the next column when approved.

[ilmoniemi]

Blocked by #83: overlapping changes to ConversationRepository.kt, FakeConversationRepositoryTest.kt, and ChannelListViewModelTest.kt. #83 is a Spotless/ktlint wiring ticket that reformats whitespace across these same files (418 + 591 lines of reformatting); shipping #93 in parallel would produce a merge conflict on every reformatted region. Will write the spec once #83 lands.