test: Details-clamp E2E 'advert links in Details stay visible and clickable' is time-dependent and flaky

[dborup]

Problem

The Playwright step "advert links in Details stay visible and clickable" (Kpa-clawbot#1122, Details clamp) fails intermittently at 1200 and 900 px:

advert link in Details is not hit-testable: {"found":true,"hitIsLink":false,"text":"R5-D4 300D Rak"}

• It failed on master aff158c7 (CI run 37274272380), and on PR branches that merged that master: feat(packets): optional Hide CONTROL packets filter (#96) #211, fix(css): raise the remaining 44 px controls to the 48 px touch-target minimum (#235) #239 and fix(ingestor): refuse a hard-linked stats tmp, and harden the #160/#161 tests (#228) #240. One of them, fix(ingestor): refuse a hard-linked stats tmp, and harden the #160/#161 tests (#228) #240, is an ingestor-only change.
• A rerun of the same master job passed, and fix(css): raise the remaining 44 px controls to the 48 px touch-target minimum (#235) #239 passed on its second attempt.
• It is therefore time-dependent: the freshened fixture changes which rows are in the default window, and so which advert link is found and what sits on top of it at its hit point. It is not a regression in any of those PRs.

This test failed before for the same reason (fixture age, Details-column wrap). PR #67 (the Details row clamp) fixed that case.

Proposed approach

• Make the test deterministic. Pin the row it inspects, e.g. a seeded advert with a known name, instead of the first match in a time-dependent window. Freeze or seed the window, so the fixture age does not change the layout.
• Before changing the test, check whether a real overlap exists for long advert names at 900–1200 px. Log the element that is actually hit (elementFromPoint) in the failure message. If it is a real layout bug, fix the CSS instead.

Acceptance

• The failure message names the element that was hit.
• The test passes 20/20 in CI-like runs at different fixture ages, e.g. freshened 0 and 2 hours earlier.
• If a real overlap is found, it is fixed with a test.

[dborup]

Fixed by #252 (merged as 310c501a).