Add /display left-rail CSS source map and permanent fidelity guard test

[WilfordGrimley]

Description

Two-part CSS-fidelity task on the /display left rail
(docs/proposals/mockups/proposal-h/SPEC-display-left-rail.md), triggered by
PR #352 shipping and then having to separately fix its own §2 density-table
regression ("Fix left-rail CSS fidelity regressions vs
SPEC-display-left-rail.md").

1. Source map. Traced every §0 (theme token) and §2 (density) binding
value to where it actually resolves in the code today. Full table lives in
the spec file's new §12 Source map addendum — summary:

Category	Finding
Most §2 padding/margin/gap values (rail header, artist line, D14 band, Select Version wrapper, unified filter fieldset, .vgrid grid, Slot Actions stack, Sources bulk row/list)	Already component-scoped inline styles or styled() rules — safe, no action needed
§0 theme tokens ($body-bg, $secondary, $primary, $success, etc.)	Intentionally global via bootswatch/superhero SCSS + Bootstrap variant classes — correct pattern, not a bug
AutofillCollapse header padding 7px 10px	This is the actual recurrence mechanism. Was a RailRoot-level .card-header{padding:7px 10px} plain descendant selector living two files away from Card.Header, clobbering Bootstrap's global card.scss rule by selector specificity instead of by scope — exactly the "pinned in a higher location than expected" pattern. Fixed: added an additive, optional headerPadding?: string prop directly on AutofillCollapse.tsx, passed at each rail call site (RailSection, SourcesAccordion). Every other caller (CardDetailedViewBody/PDFGenerator/JumpToVersion/CardResultSet/GridSelectorFilters) omits the prop and is byte-for-byte unchanged.
AutofillCollapse header background	Found and fixed a real, independent drift while in the file: hardcoded #4E5D6B was one hex digit off the actual $secondary token (#4e5d6c) — never sourced from the theme at all. Corrected to the exact value.
Container gutter under AutofillCollapse body pad={2}	Judgment item, not re-verified this pass — the spec's own §2 row already flagged "verify no double gutter"; still resolves through Bootstrap's global --bs-gutter-x, listed as an open risk in the addendum, not fixed here
D14 band's literal hex colors (.d14/.seticon/.score/.statepill)	Accepted trade-off, not fixed: spec §0 deliberately calls for literal-extracted values over CSS variables, so these are immune to Bootstrap clobber but won't auto-track a future theme-token edit — noted, not addressed (would need a larger CSS-custom-property refactor)

2. Permanent fidelity guard. Added frontend/tests/DisplayLeftRailFidelity.spec.ts
— asserts real getComputedStyle values (toHaveCSS, not class names or
source text) against the spec's own literal §0/§2 binding values, reusing
SelectVersionSection.spec.ts's navigation flow (now hoisted into
test-utils.ts's openSelectVersionSection so both files share it, rather
than one spec importing from another). Comments in the new spec point back
to SPEC-display-left-rail.md as source of truth so future edits update both
deliberately.

Checklist

• I have installed pre-commit and installed the hooks with pre-commit install before creating any commits.
• I have updated any related tests for code I modified or added new tests where appropriate — new DisplayLeftRailFidelity.spec.ts.
• I have manually tested my changes as follows:
  • Ran the new fidelity spec twice locally (both green): npx playwright test tests/DisplayLeftRailFidelity.spec.ts --project=chromium.
  • Ran SelectVersionSection.spec.ts (7/7 green) to confirm the shared-helper refactor didn't regress issue Select Version section build (blocked on PR #116) #167's own coverage.
  • Ran PDFGenerator.spec.ts and GridSelectorModal.spec.ts alone (both fully green) to confirm the other AutofillCollapse callers are unaffected by the new headerPadding prop.
  • Full jest suite: 568/568 passed. tsc --noEmit: clean.
  • npx prettier@2.7.1 --check on all changed files: clean.
  • Note: DisplayPage.spec.ts/DisplaySlotStates.spec.ts show occasional timing flakes under heavy worker parallelism in this sandbox — reproduced identically on pristine master with the same worker count (not caused by this change); each passes cleanly alone or with --workers=1.
• I have updated any relevant documentation or created new documentation where appropriate — SPEC-display-left-rail.md's new §12 Source map addendum, edited in place.