docs: complete the Docusaurus (pnpm monorepo) case study

We want a verified case study of CVE Lite CLI on Docusaurus (Meta's documentation framework, a large pnpm monorepo with ~2,590 resolved packages). An earlier draft (PR #766, now closed) got the scan and most of the case-study structure in place but stalled before the measured-results step, so this is a good pickup for a first-time contributor.

Starting point: the closed PR #766 is a useful reference for the scan output and the overall structure - you can build on it rather than starting from scratch.

Use an existing study such as analog.md as the template and follow the standard case-study section order.

What's needed to finish

1. Before vs After table with measured rows (the core of the study). It needs the baseline row plus one measured row per fix-command group. Clone Docusaurus at a pinned commit, apply the first fix-command group, rescan with cve-lite . --verbose, record the row (findings / severity / direct / transitive / command groups), then apply the next group and repeat. Run git checkout -- . to restore between passes. This is what actually proves the fix commands work rather than just claiming it.
2. Fix the "Command groups" column direction so it starts at the full count (4) and decrements to 0 as each group is applied.
3. Single linked project line at the top (**Project:** [Docusaurus](https://github.com/facebook/docusaurus) - ...) matching the analog.md / nestjs.md pattern, rather than a separate unlinked Project line plus a Repository line.

Notes

• Follow the full section order: logo (centered), Summary, What this case study demonstrates, Comparison Note vs npm/pnpm audit, Before vs After, Fix Journey, Why this matters, Scan command, Remaining risk, Baseline findings, Want your project reviewed?
• No em dashes; use proper markdown tables (no ASCII/box-drawing tables).
• The sidebar and case-studies index wiring are handled by the maintainer after merge, so you do not need to touch those.

Scan reference: CVE Lite CLI found 14 unique vulnerable packages (0 critical, 6 high, 7 medium, 1 low), 1 direct / 13 transitive, 20 CVEs matched, 4 fix-command groups covering 7 of 14 findings.