fix: replace the traced infinity mark with the real artwork; add §4 Offerings

Q32 — the traced mark was a WRONG SHAPE and had shipped. Pouya compared
it against the master and rejected it. Two grounds reproduce from the
path and are verified here: all four cubic branches meet the origin at
exactly 90 degrees, so the loops are tangent rather than crossing and at
stroke-width 28 render as two kissing circles (signed crossing number 0;
the strokes fuse across 61% of the mark's height at 2rem); and the
master's ink bbox is 2668x1704 = 1.5657:1. The path is deleted, not kept
as a fallback.

Pouya's 1.23:1 figure is reconciled rather than left dangling: it is the
bounding box of the path's COORDINATES, not the curve. Control points sit
at y +/-160 where the curve reaches +/-120, so the hull is 400x320 and
with stroke 428x348 = 1.2299. A trap rather than a slip — x is monotone,
so the control points give the right width and a 33% inflated height, and
the "does the width look right" check passes.

The real artwork is now in the repo: master, tight crop (the render
source, so the file's aspect ratio IS the mark's), full lockup, and the
SVG. InfinityMark renders AVIF/WebP; a Retina device takes 3,063 B.
Favicons regenerated; favicon.svg deleted.

Q33/Q36 — Pouya accepts arbitration appointments now. §4 gains an
Offerings category: competence for an offering, permission for a
credential, with an explicit boundary so it cannot become a route around
D13. The masthead tagline is restored, and the footer designation strip
now carries "Q.Arb — commenced August 2026" so §4's paired-disclosure
condition is actually met on every page rather than only asserted.

Two conventions added to CLAUDE.md, both earned this session: anything a
spec makes a claim about must be reachable from the repo (R14 — the
traced mark survived two review passes because the artwork was not here
to compare against); and a command that did not run is not evidence of
absence (`timeout` is not installed on macOS, so four Drive reads never
executed and were reported as an empty directory).

Reviews: claims-auditor FAIL/13 and adversarial-reviewer 2 blocking, all
resolved. The severe one was self-inflicted — `flex: none` landed on the
<img> while <Picture>'s <picture> wrapper is the flex item, so the logo
compressed to 28.5x32 at 1024px with seven nav items. The page-level
overflow check passed throughout because the brand block absorbed the
deficit by crushing the mark. Harness now asserts rendered aspect ratio.

Opened: Q38, Q39. Closed: Q32, Q33, Q36. Narrowed: Q35. Added: R13, R14.
AGENTS.md entry (v) carries a RESUME HERE section.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012XquaEq4BgWMCwUqLEyNkF

src/styles/tokens.css

@@ -112,13 +112,16 @@
 
   /* --- Layout ------------------------------------------------------------ */
 
-  /* Sticky-header height at >= 64rem, where the header IS sticky. global.css
+  /* Sticky-header height at >= 66rem, where the header IS sticky. global.css
      drives `scroll-padding-top` off this, so the skip link does not drop the
      reader behind the header. If SiteHeader's padding or nav sizing changes,
      re-measure and change this with it — one fact living in two files.
      [measured 2026-08-26 — headless Chrome at 1024/1100/1280/1440px, with six
-      nav items and with a seventh injected: 77px at every one]               */
-  --header-h: 4.8125rem;    /* 77 — measured, not chosen */
+      nav items and with a seventh injected. 81px at every one: 32 padding + 48
+      reserved brand block + the 1px bottom border, which is easy to forget and
+      is why this is measured rather than added up. The brand reserves 48px so
+      the height does not change when the tagline appears at 76rem]            */
+  --header-h: 5.0625rem;    /* 81 — measured, not chosen */
 
   --width-content: 80rem;   /* 1280 */
   --width-wide:    90rem;   /* 1440 */