fix: retain srcset width for duplicate image URLs

[linjiaye929]

Summary

An image such as <img src="photo.jpg" srcset="photo.jpg 640w" alt="Photo"> loses its declared width: the src entry is added first, and URL deduplication discards the richer srcset entry.

Keep one variant per URL, but fill its missing width from a later source. Existing widths, variant order, and filtering stay unchanged. This preserves the srcset width descriptor; it does not measure image dimensions.

List of files changed and why

• content_scraping_strategy.py: retain metadata while deduplicating variants.
• tests/test_image_variant_width.py: cover public scraping and serialized results.

How Has This Been Tested?

45 focused tests passed: 14 input scenarios, one legacy-alias smoke test, and 30 existing scraper regressions. Cases include distinct URLs, first-present width precedence, zero/invalid descriptors, and lazy-loading attributes. No network or model service was used; the full browser/integration suite was not run.

New-test Black and diff checks pass. Source-range Black reports existing whitespace differences also present on the baseline; unrelated formatting is unchanged.

AI assistance was used for investigation, implementation, tests, and review.

Checklist

• Added regression tests and checked focused tests locally.
• Reviewed the patch.
• Full browser/integration suite run.
• Documentation change (no API change).

Hosted status (2026-10-16 07:08 UTC): no code/test checks are reported for this head. The Discord notification workflow requires maintainer action. Validation above is local.