fix(1688): extract detail images from shadow DOM with lazy-render scrolling

[HARRY-BEAR]

Problem

The 1688 assets / download commands return detail_count: 0 on real offer pages, even though the page contains 7–16 detail images. Only main images and videos were extracted.

Root cause

Two compounding issues:

1. Shadow DOM pierce failure (structural). The product detail section lives inside the shadow root of a custom element (v-detail-e, class html-description). The old code queried selectors like .html-description img through queryAllDeep, which runs the same selector inside each shadowRoot — but .html-description is the host element in light DOM, so the selector can never match inside the shadow root. The detail group was always empty.

2. Lazy rendering (timing). The detail content renders only after the user scrolls near it. The old autoScroll({ times: 3, delayMs: 400 }) is not enough for long pages (~13k–18k px), so the lazy images never materialized before extraction.

Fix

• Collect all img / source elements while walking shadow roots, then decide membership by walking the shadow host chain with closest() against the known detail containers (.html-description, .de-description-detail, #detailContentContainer, .desc-lazyload-container).
• Scroll more aggressively (autoScroll 6+4 × 500ms), then scrollIntoView the detail container and settle for 3s before extraction.

Verification

Real-browser results before → after:

offer	detail before	detail after
997412525461	0	16
954578900038	0	13
871254382620	0	7

download end-to-end lands *_detail_NN.jpg files; main/video counts unchanged (no regression).

Added a jsdom regression test (assets.test.js) covering:

• images inside the v-detail-e shadow root → detected as detail
• light-DOM main gallery images → not detected as detail
• light-DOM plain-class detail containers → detected as detail

All 15 tests in clis/1688 pass; npx tsc --noEmit is clean.