Normalize /display left rail dividers to #16202b (O1 fidelity round)

[WilfordGrimley]

Description

Corrected SPEC-display-left-rail.md fidelity round (Quorra's owner-approved corrected spec + mockup). Most of the corrected round's binding table was already shipped in PR #352 (Sources disclosure, unified Frame+Treatment filter, preferred-sources pin with the #353 seam, artist support button, the D14 Y/N consensus treatment, the buttons-look-like-buttons sweep) — verified against the corrected spec and left untouched where it already matched. This PR ships the one genuinely new binding change (O1) plus the doc/test/mockup sync the corrected round requires.

Per-item mapping to the corrected spec:

1. Sources disclosure (~247 rows) — already shipped (SourcesAccordion.tsx): summary count, pinned-source chips, filter input, bulk enable/disable/invert, 38px toggle rows. Verified matches §D.1, no changes needed.
2. Frame + Treatment unified filter block — already shipped (SelectVersionResults.tsx's .ufilter fieldset, Border its own row above). Only its border color changed this round (see O1).
3. Preferred-sources pin ★ (38×38) — already shipped, persists to localStorage (getLocalStoragePinnedSourcePks/setLocalStoragePinnedSourcePks), disabled #353 "Save as my defaults" seam present. Added a new Playwright test asserting the pin survives a real reload.
4. Artist support link — already shipped as btn-outline-primary "Support on MTG Artist Connection ↗" (ArtistSection.tsx).
5. Buttons-look-like-buttons sweep — already shipped across the rail's offender list (Slot Actions, D14 ✗, Sources bulk/pin/save-defaults, artist link).
6. O1 — divider normalization to #16202b — implemented this PR. .rail-head/.artist-line/.sources (previously the ambiguous Bootstrap .border-bottom utility), the Select Version wrapper (previously no boundary at all), and the unified filter fieldset + its internal Frame↔Treatment divider (previously an unthemed rgba(0,0,0,.22)) all now carry the explicit #16202b hairline .d14 already used.
7. Y/N consensus treatment (§C) — already shipped exactly as specced: confirmed = green ✓ + "Confirmed" pill (no number); suggested = score/qualitative pill + amber "N% confident"/"Suggested"; ✗ not this printing is a real vote that stays visible at opacity:.6 on a confirmed printing (owner answer Promote PringlePrints to a full ordering tab #2). No standalone Y/N block anywhere in the rail.

Docs updated in place (no dated appendix): docs/proposals/mockups/proposal-h/SPEC-display-left-rail.md rewritten to the corrected binding tables (§A–§J, including a refreshed source-map addendum); docs/proposals/mockups/proposal-h/display-left-rail-mockup.html replaced with the corrected mockup; docs/features/display-left-rail.md gained an "O1 — divider normalization" section.

frontend/tests/DisplayLeftRailFidelity.spec.ts updated to assert the new #16202b = rgb(22, 32, 43) border values (comment-linked to the spec's §D.1/O1), and to flag the one spec row deliberately NOT implemented (see Deviations).

Deviations from spec

• .btn-sm sizing (14px / 4px 8px, spec §A item 2 / §D.1's final row) — NOT implemented. The corrected spec's own §A explicitly flags this specific row as "needs owner sign-off" (denser-vs-standard tradeoff) separately from O1's already-confirmed divider normalization. It directly conflicts with a more specific, already-shipped owner directive on the exact same controls ("the buttons are too big" — CompactButton/CompactToggleButton/CompactLinkButton/TreatmentChip, currently ~12px/0.2rem 0.5rem). Left as shipped pending an explicit owner call; flagged in the spec (§A/§H "O5"), the source-map addendum (§J), the fidelity test's own comment at the funnel-filters-toggle assertion, and docs/features/display-left-rail.md.
• Everything else in the corrected spec's binding table (§D.0/§D.1) was already shipped via PR Shrink Select Version rail's medium/hero card tile widths #352 and verified to match on inspection — no other deviations.

Verification

• DisplayLeftRailFidelity.spec.ts — 2/2 passed (asserts the new #16202b border values).
• DisplayPage.spec.ts — 32/32 passed (includes the existing Sources/D14 coverage plus the new pin-persists-across-reload test).
• SelectVersionSection.spec.ts — 6/6 passed. DisplaySlotStates.spec.ts — 4/4 passed in isolation (one test flaked once under 4-worker CPU contention in a combined run, unrelated to this change — passed cleanly re-run alone, matching this repo's documented CI shard-contention pattern).
• Full jest — 568/568 passed.
• tsc --noEmit — clean.
• npx prettier@2.7.1 --check — clean on all changed files.
• Screenshots (desktop 1400px + phone 390px) captured locally via a throwaway Playwright script and inspected — rail composition, divider hairlines, Sources accordion (expanded + collapsed), D14, and the unified filter block all render as expected; not committed (gitignored test-results/).
• Deferred: full visual re-baseline against the mockup's own shot.js screenshots (mockup-side tooling, not part of this repo) — the CSS-fidelity guard's getComputedStyle assertions are the binding check for this repo per its own established convention.

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.
• I have manually tested my changes as follows:
  • Ran the full targeted Playwright suite (fidelity guard + DisplayPage + SelectVersionSection + DisplaySlotStates) plus full jest, tsc --noEmit, and prettier - all listed above.
  • Captured and visually reviewed desktop (1400px) and phone (390px) screenshots of the rail (idle, Sources expanded, D14 suggested state).
• I have updated any relevant documentation or created new documentation where appropriate.
  • docs/proposals/mockups/proposal-h/SPEC-display-left-rail.md (rewritten in place), docs/proposals/mockups/proposal-h/display-left-rail-mockup.html (replaced), docs/features/display-left-rail.md (O1 section added).
  • Merge-time checklist: wiki: unified-page docs need the new Sources/filters/pin controls documented.

Do not merge - for owner review.

[WilfordGrimley]

Follow-up round: machine-diff fix pass

A throwaway computed-style diff against the corrected mockup (bundle-401) caught 63 property mismatches beyond the O1 divider round. Fixed every unambiguous one this round (mostly Bootstrap body-default font-size fall-throughs on bespoke rail classnames, plus the Source row's own border-bottom and the tile corner tag's font-size/alpha), all component-scoped per the #400 rule.

Per explicit owner ruling, also resolved the three rows previously held back pending sign-off:

• AutofillCollapse header background reverts to #4E5D6B (PR Add /display left-rail CSS source map and permanent fidelity guard test #400 mis-corrected this to #4e5d6c — deliberately distinct token, not a typo)
• Source toggle restyled (scoped to a new rail-source-toggle class, ~10 other react-bootstrap-toggle mounts sitewide unaffected) from the library's stock sliding switch into the mockup's static two-cell segmented look
• Filters disclosure toggle returns to real Bootstrap sm metrics (14px/4px 8px), superseding the earlier "buttons are too big" shrink for that one control only

Re-measured after the fix: 63 → 44 raw mismatches (re-diff via the same diff.js tool against a fresh built-page extraction). Of the remaining 44: most are pre-existing, already-documented caveats (fixture/measurement limitations — suggested-variant Frame chip, RequestedPrintingBadge wrapper-vs-badge, the confirm-ribbon's fundamentally different implementation, tile density-tier sizing) or deliberately-deferred items flagged in the spec doc (Demoted body font-size — blast radius across four unmeasured sections not verified; the Sources/Slot-Actions header's own border-bottom, left at Bootstrap's stock Card border once the stray border-light utility was removed, since no spec pins a literal value there). A couple of the remaining "Source toggle"/"Source pin star" rows are artifacts of my own diagnostic re-extraction script's fixture state (e.g. comparing an unpinned built row against the mockup's demo data, which hardcodes its first row as pinned) — not real defects; the toggle's visible cells were verified pixel-correct via screenshot and the fidelity guard's own getComputedStyle assertions.

DisplayLeftRailFidelity.spec.ts asserts every value fixed this round. Full verification (tsc, prettier, eslint, jest 568/568, and the DisplayLeftRailFidelity/DisplayPage/SelectVersionSection/DisplaySlotStates/PDFGenerator/GridSelectorModal Playwright suites, 76 tests) all green — see commit 980b6cb1.