test(desktop): await thread scroll anchor

[cameronhotchkies]

Why

The focus/split E2E test could capture the thread root before its programmatic middle-thread scroll had settled, then incorrectly report a scroll-restoration failure.

What

• Poll until the requested middle-thread scroll position is applied
• Require the captured anchor to intersect the thread viewport and differ from the root
• Preserve the existing focus-to-split-to-focus viewport assertions

Risk Assessment

Low — test-only synchronization change with no production behavior changes.

References

• Original failure: https://github.com/block/buzz/actions/runs/30231271427/job/89870533541
• Buzz thread: buzz://message?channel=12dd513d-45fd-48ff-80ac-8596d2fcc9d3&id=87ce6024b4bf74bfac2fa75d9f7bbbcc8f8fe2df460afe534152c495929f51ba
• Reproduced confidence: 20 consecutive targeted passes, full spec pass, just desktop-ci, and just ci

Generated with Codex

[atishpatel]

Approving — test-only synchronization fix, verified locally.

What I checked

• Reviewed the diff: topVisibleMessageId → scrollToMiddleVisibleMessage, which (a) sets the 40% scroll inside the poll and re-checks element.scrollTop landed within 1px in the same synchronous evaluate, (b) requires the captured row to intersect the viewport on both edges, (c) rejects the thread root. That is exactly the failure mode in the linked run: the anchor was captured at scrollTop === 0, so the root became the anchor and the later restoration assertion legitimately failed on an illegitimate anchor.
• Not masking a product bug: the position check is synchronous with the assignment, so the anchor is genuinely a ~40% row. If restoration were broken (or the app snapped back to bottom asynchronously afterward), the preserved toBeInViewport() assertions after focus→split→focus would still fail. The three original viewport/highlight assertions are untouched.
• Rebased the diff cleanly onto origin/main (96ae141) in a scratch worktree — the branch is 159 commits behind but thread-focus-mode.spec.ts has no conflicting change on main since cca88390.
• pnpm exec tsc --noEmit: clean. pnpm check (biome + file-size/px-text/pubkey guards): clean, no new findings.
• playwright test --project=smoke thread-focus-mode.spec.ts --repeat-each=8: 16/16 passed, no flaky, on the rebased tree.

Nits, non-blocking

• expect.poll with a mutating callback is slightly unusual (the poll body performs the scroll rather than just observing it). It's correct here — the retry is the re-scroll — but a one-line comment saying so would help the next reader.
• The anchorId closure write plus the post-poll if (!anchorId) throw is redundant with the poll predicate; harmless.

Nothing blocking. Please rebase/let the merge queue bring it up to date before landing.