fix(capture): distinguish SVG namespaces from attribute values

[user-github-me]

Capturing an inline SVG whose label contains text such as xmlns=demo saves a logo that fails to load as an image: toStandaloneSvg mistakes the label text for a namespace declaration and omits the real SVG namespace. The same regex handling also misreads xlink-looking labels and quoted tag delimiters.

Read namespace declarations and xlink usage from parsed attribute names, then insert the required declarations into the original source. Preserve artwork and label bytes, and leave already-declared SVGs unchanged. This uses the CLI's existing linkedom dependency.

Validation: seven added regressions fail on main, including an actual asset download; all 374 related capture/command tests pass with the fix. Eight Chrome 152 cases run through the real token extractor and asset downloader, then load each saved SVG as an image and verify its dimensions, label, and expected pixel. Existing namespace cases retain byte-identical source. CLI build/typecheck, root lint, formatting, comments, and commit hooks passed.

[jrusso1020]

Thanks for this. The failure needs an SVG attribute value that literally contains xmlns=, which captured sites don't produce in practice, and the quoted > case doesn't occur either because the browser escapes it in outerHTML. Given that, we'd prefer to keep the simpler regex path rather than parse every captured SVG into a DOM. — Rames