test(vr): regenerate __screenshots__ baselines on Linux (CI Playwright image)

The committed baselines were macOS-rendered (different font binaries,
glyph metrics, anti-aliasing), so the CI Visual Regression Matrix job
running in the pinned `mcr.microsoft.com/playwright:v1.60.0-jammy@sha256:
e1529a04...` image diff'd every cell against them. Result: 48/48 red on
CI for the past dozen+ pushes despite local being clean.

Regenerated all 8 Vue-keyed baselines by running the canonical CI image
locally via docker:

  docker run --rm -v $PWD:/workspace -w /workspace \
    mcr.microsoft.com/playwright:v1.60.0-jammy \
    bash -c 'npm i -g pnpm@10 && CI=true pnpm install --frozen-lockfile \
      && pnpm turbo run build \
      && cd tests/visual-regression \
      && pnpm exec playwright test --grep "· vue" -u \
      && pnpm exec playwright test'

Vue baselines were captured first (matches `D-10 baseline keyed by
example` rule), then the full 48-cell matrix was re-run against them:
48/48 pass inside the CI image.

Per D-12 the pinned-digest container is the single authoritative
screenshot environment; macOS local runs are advisory only.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: serpentblade