feat(data): seed FakeConversationRepository with 3 fake channels

[ilmoniemi]

User Story

As a developer building Phase 0 channel-list UI, I want FakeConversationRepository to start with three pre-populated channels so that composables have realistic data to render without each call site having to call createDiscussion + promote in setup.

Context

Phase 0 UI work (channel list screen and downstream) needs fixture data. FakeConversationRepository currently initializes empty (delivered by #4). Three seeded channels with distinct cwd / currentSessionId / lastUsedAt give the channel-list screen something to render against until the Phase 4 backend ships.

The parallel discussions seed (#6) handles isPromoted = false entries; this ticket must leave discussions state untouched.

Acceptance Criteria

• Three channels (isPromoted = true) populate FakeConversationRepository's initial state on construction — e.g. "Pyrycode Mobile", "Joi Pilates", "Personal" (or similar).
• Each seeded channel has a distinct cwd, a stable currentSessionId, and a distinct lastUsedAt.
• observeConversations(Channels) emits all three on first collection.
• observeConversations(Discussions) emits empty (feat(data): seed FakeConversationRepository with 2 fake discussions #6 runs in parallel and is out of scope here).
• A new test asserts the three are returned and ordered by lastUsedAt descending.

Technical Notes

• The existing observeConversations_emitsEmpty_initially_for_all_filters test in FakeConversationRepositoryTest will fail once initial state is non-empty — adjust or replace it as part of this change (count goes from 0 to 3 for Channels; Discussions stays at 0).
• observeConversations does not currently sort. The ordering AC implies adding a sortedByDescending { it.lastUsedAt } projection inside the existing state.map { ... } block.
• Use kotlinx.datetime.Clock.System.now() offset by kotlin.time.Duration deltas (or Instant.fromEpochSeconds) so each seed gets a deterministic, distinct lastUsedAt.

Size Estimate

XS — three seed entries in the state initializer, one short sortedByDescending, one new test, one adjusted existing test.

Depends on

• feat(data): scaffold FakeConversationRepository (empty state) #4 (skeleton, closed/merged)

[ilmoniemi]

🤖 Product Owner — creates structured issues

po agent has completed work on this ticket.

Agent output (click to expand)

Ticket #5 refined. Size already correctly labeled `size:xs`; no split needed (single concern, 5 AC at the limit, single file + test changes).

Summary of changes to the body:
- Reframed "Goal" as a proper **User Story** tying the seed data to Phase 0 channel-list UI work.
- Added **Context** explaining the dependency on #4 (now closed/merged), the Phase 4 motivation, and the boundary with the parallel discussions seed (#6).
- Preserved all five original acceptance criteria, lightly tightened ("populate ... initial state on construction", "first collection") to remove ambiguity.
- Added **Technical Notes** flagging two non-obvious things the architect needs:
  1. The existing `observeConversations_emitsEmpty_initially_for_all_filters` test will break — must be adjusted.
  2. `observeConversations` doesn't currently sort, so AC #5 (`lastUsedAt` desc ordering) implies adding `sortedByDescending` inside the existing `state.map` block.
- Added **Size Estimate** narrative justifying XS (already correctly labeled).

Not applying `security-sensitive` — pure in-memory test fixture, no trust boundaries. Dispatcher will auto-add `ready:po` and advance to In Architecture.

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