fix(budget-sources): add horizontal divider between work item groups (#1309)

[steilerDev]

Summary

• Add a horizontal border-top between consecutive .parentItemBlock siblings in SourceBudgetLinePanel so work-item groups are visually separated in the Budget > Sources list.
• Uses adjacent-sibling selector .parentItemBlock + .parentItemBlock with --color-border + --spacing-2 tokens (auto-themed for dark mode, no trailing divider after the last group).
• Unit tests document the DOM contract the CSS selector depends on; E2E tests assert the computed border-top across desktop/tablet/mobile viewports.

Fixes #1309

Test plan

• Unit tests: structural DOM contract in `SourceBudgetLinePanel.test.tsx`
• E2E tests: computed-style assertions in `e2e/tests/budget/budget-source-lines.spec.ts` (`@responsive`)
• Stylelint: tokens only, no hardcoded colors/spacing
• Quality Gates CI green

Co-Authored-By: Claude Opus 4.6 noreply@anthropic.com

[github-actions]

Thank you for your submission! We require all contributors to sign our Contributor License Agreement before we can accept your contribution.

To sign, please comment on this PR with:
I have read the CLA Document and I hereby sign the CLA

---

I have read the CLA Document and I hereby sign the CLA

---

Frank Steiler seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[steilerDev]

[product-architect] Architecture compliance confirmed (posting as comment; cannot self-approve). CSS uses design tokens (--color-border, --spacing-2) — no hardcoded values. The adjacent-sibling combinator .parentItemBlock + .parentItemBlock is the idiomatic solution for "separator between, never after last" and keeps the markup untouched. File locations and naming match conventions: component CSS module co-located with component, unit test .test.tsx co-located, E2E spec under e2e/tests/budget/ with @responsive tag for cross-viewport coverage. The JSDOM workaround in the unit tests is correctly documented — asserting the DOM contract the selector depends on and deferring computed-style verification to E2E is the right split. No schema, API contract, or ADR impact. No concerns.

[steilerDev]

[ux-designer] Design review: APPROVED (posted as comment — cannot approve own PR via GraphQL).

Token adherence. Both values use the correct semantic tokens: var(--color-border) (gray-200 light / slate-500 dark, with a full [data-theme="dark"] override in tokens.css) and var(--spacing-2) (8px). No raw values.

Visual hierarchy. The three-level hierarchy holds cleanly: 2px --color-border-strong left accent (group identity) > 1px --color-border horizontal rule (inter-group separation) > 1px --color-border lineRow bottom divider (intra-group rows). Using the lighter --color-border rather than --color-border-strong for the horizontal rule preserves that hierarchy — the left accent continues to dominate. The --spacing-2 (8px) breathing room on both sides matches the cadence already established by .lineRow padding. The adjacent-sibling margin-top: var(--spacing-2) cleanly overrides the base .parentItemBlock top margin (--spacing-0-5) for all blocks after the first — no unintended stacking.

Dark mode. --color-border is redefined in the [data-theme="dark"] layer in tokens.css (--color-slate-500), so the rule switches automatically in both themes with no additional overrides needed.

Accessibility. Purely decorative separator with no semantic meaning — no ARIA role, label, or contrast obligation applies.

Responsive. Layout-agnostic; renders correctly at all viewport widths.

Tests. The unit tests correctly target the DOM contract (sibling count + shared parent element) rather than computed styles — appropriate given JSDOM's inability to resolve CSS Modules. The E2E tests verify borderTopStyle === 'solid' and borderTopWidth !== '0px' on the second block in a real browser, with both positive (two groups) and negative (one group) cases covered.