fix: correct text contrast in download section for dark mode

[swastik-basu]

Summary

Fixes unreadable text (white text on white background) in the "Try it now" download
section when the site is viewed in dark mode.

Closes #2920

Root cause

LandingPageDownload.client.vue sets a fixed light background
(bg-primary-50 bg-opacity-50) but never sets an explicit text color. The
codebase relies entirely on Nuxt UI's automatic light/dark semantic tokens
(e.g. text-highlighted) with no manual dark: overrides anywhere — so this
section's text was inheriting a mode-adaptive color intended to pair with a
mode-adaptive background. Since this section's background is fixed instead,
dark mode produced light text on a light background.

Fix

Added text-primary-900 to the section's content wrapper.
--color-primary-900 is already defined in assets/css/main.css (existing
brand primary scale) — no new colors, tokens, or palette changes introduced.
A mode-adaptive semantic token (e.g. text-highlighted) was intentionally
not used here, since it would reintroduce the same mismatch against a
background that doesn't adapt with color mode.

Testing

• Verified in both light mode and emulated prefers-color-scheme: dark
• Checked Storybook story (LandingPageDownload.stories.vue)
• pnpm test passes

Screenshots

[white text, unreadable] [dark text, readable] |

[swastik-basu]

Hi! The CI workflow is currently not running because the PR is not labeled safe to test. Could someone with the required permissions please mark it safe to test? Thanks!

[ThiloteE]

Does this also address the "Also available for"?

[swastik-basu]

Yes — "Also available for" and the OS-name text in all three branches (Windows/Mac/Linux) are plain text nodes with no color class of their own, so they inherit text-primary-900 from the shared wrapper div the fix was applied to (line 7). It's a single change covering all OS variants, not just Windows.

[swastik-basu]

Also — the UI Tests check is showing "1 change must be accepted as baseline," which I believe is just the visual regression tool correctly flagging the intentional contrast fix as a new diff. Let me know if there's anything I need to do on my end, or if a maintainer needs to accept it as the new baseline.

[koppor]

Follow-up to #2924

[swastik-basu]

Follow-up to #2924

Thanks for the reference. I looked at #2924 — noticed it introduced new CSS variables (--surface-*) and manual dark: overrides, which I believe is why it was later reverted from main. This PR takes a narrower approach: it reuses primary-900, already defined in the existing palette, with a single-line change and no new tokens or dark: overrides.