Visual regression baseline — screenshot diffs on PR

Summary

Establish a visual regression testing workflow using Playwright screenshots as the baseline.

Approach

After the Tailwind migration stabilizes, capture screenshot baselines for each page × viewport and diff against them on every PR to catch unintended visual changes.

Tasks

• Choose a visual diff tool: pytest-playwright + pixelmatch, or percy.io, or BackstopJS
• Capture baseline screenshots (desktop/tablet/mobile) for: login, dashboard, accounts, ledgers, settings, profile
• Commit baselines to tests/ui_functional/snapshots/
• Add CI step: diff screenshots vs baselines, fail on > 0.5% pixel difference
• Add --update-snapshots flag workflow for intentional UI changes
• Document snapshot update process in CONTRIBUTING.md

Acceptance criteria

• Unintentional CSS regressions caught on PR
• Snapshot update is a deliberate, documented action

Part of epic #299. Closes #291 (visual regression testing).

Start by reviewing the existing UI tests and the target path tests/ui_functional/snapshots/, then confirm the Tailwind migration is stable. Resolve the visual diff tool choice, capture baselines for the six listed pages across desktop, tablet, and mobile, and wire the PR comparison and deliberate snapshot-update workflow. Document the update process in CONTRIBUTING.md and verify that the acceptance criteria are met.