Landing QA fixes + English content quality cleanup

[haksungjang]

Overview

The first batch covering ① (landing QA) and ② (P-growth). Code-level QA fixes for the landing page just deployed, plus quality cleanup of the English content.

① Landing QA

With no browser capture tooling available, I performed a code-level inspection (contrast-ratio calculation and robustness checks):

• Found and fixed: the hero titleAccent used color: transparent, so the accent text could disappear in browsers that don't support background-clip: text → added a fallback color (#8ab4f8).
• Contrast ratio (WCAG AA) calculations pass: meta text (rgba 0.62 on white) ≈ 6:1, with the title (white) and subtitle (0.82) above that; the primary CTA (deep blue on white) and secondary CTA (white on dark) are high-contrast. The hero mockup text is aria-hidden, so it's exempt from accessibility requirements.
• Responsive behavior (CTA stacks vertically and the mockup transform is removed at ≤640px) and light/dark theme handling both work correctly.

② English content quality (P-growth English parity, round 1)

Cleaned up 8 places in the English content where machine-translation artifacts remained (sentences broken by ,,), rewriting them in natural English:

• 00-overview/supply-chain (XZ Utils and CRA paragraphs), 05-tools/sbom-generation (4 places), 05-tools/vulnerability, 06-training.

Verification

• npm run build (ko/en) SUCCESS, 0 broken links.
• verify.sh 12/12 PASS.

Beyond ,,, the English text still has other residual MT traces such as comma spacing and awkward word order (a separate broader pass is recommended). This PR prioritizes the most prominent sentence breaks.
The rest of P-growth (term hover tooltips, authority signals, SEO metadata) will follow in subsequent PRs.