Take the prose reset and footer from @robertblust/design v0.4.0

[robertblust]

Summary

blust.ch adopts the prose reset and prose footer fences from @robertblust/design@0.4.0
(tag-pinned via github:robertblust/design#v0.4.0), taking the plain footer variant — this
site is the one the other two credit, so it carries no footer .credit rules. header contract
also moves v2 → v3 in this release (one word in its own comment), which is expected on all six
pages and not something this PR causes.

All six prose pages now carry both new fences: index.html, talks/index.html,
privacy/index.html, ideas/index.html, principles/index.html, model/index.html. Deck pages
are untouched.

What actually rendered differently

privacy/index.html was synced first and read as the gate: beyond the marker lines and the
header's v2→v3 word, its reset and footer bodies did not change at all — it was already one of
the pages carrying the package's majority form.

On the other five pages, syncing surfaced real byte-level drift the package now corrects:

Body font stack loses a redundant fallback (index.html, ideas/index.html):

-    font-family:"Instrument Sans", ui-sans-serif, system-ui, -apple-system, "Segoe UI", sans-serif;
+    font-family:"Instrument Sans", ui-sans-serif, system-ui, -apple-system, sans-serif;

"Segoe UI" is a Windows-only face already covered by sans-serif; it added nothing these two
pages' siblings didn't already omit.

Footer font stack loses a redundant fallback (index.html, talks/index.html):

-         font-family:"Plex Mono", ui-monospace, monospace; font-size:.76rem;
-         letter-spacing:.08em; color:var(--dim)}
+         font-family:"Plex Mono",monospace; font-size:.76rem; letter-spacing:.08em; color:var(--dim)}

index.html gains the .mono rule — it had none before, which meant nothing on this page
could correctly mark a value as data:

+  .mono{font-family:"Plex Mono", ui-monospace, "SF Mono", Menlo, monospace}

.brand svg{color:var(--c-mid)} added for consistency, not a fix, on the three pages that
lacked it (index.html, ideas/index.html, talks/index.html). This site's brand mark is
<rect class="plate"> plus <text class="rb">, filled directly by .brand .plate{fill:...} and
.brand .rb{fill:var(--c-mid)} — no currentColor, no stroke, anywhere in it. Setting color
on the <svg> changes nothing rendered here; it was added by hand (this sits outside the fence
by design, since each site's mark carries its own colours) to match the ten pages across
companygraph.io and guestgraph.io where the identical rule is load-bearing, because their marks
do use stroke="currentColor":

-  .brand svg{width:28px; height:28px; flex:0 0 auto}
+  .brand svg{width:28px; height:28px; color:var(--c-mid); flex:0 0 auto}

Confirmed harmless, not by reading the diff — .brand .rb{fill:var(--c-mid)} sets the mark's
colour directly and is unaffected by the svg's color property either way, on all four
affected/gate pages (/, /ideas/, /talks/, /privacy/).

One consequence beyond the brief's checklist, verified harmless: index.html,
talks/index.html and ideas/index.html had only 3 footer rules (no
footer a[aria-current="page"]{color:var(--c-firm)}) before this change — the shared block
always ships that 4th rule, so syncing adds it. None of those three pages currently marks any
footer link aria-current="page", so the new rule is inert today; it's normalization ahead of
future need, not a visible change.

Verification

• npm run verify — all checks pass (all 8 pages, sitemap, favicon, robots.txt).
• npm run design:check — ✓ 7 file(s) and 35 fence(s) match @robertblust/design, exit 0.
• npm run og / npm run og:check — all 8 cards re-rendered; og:check reports every card
  matches the page it renders. (Only each page's og.sha changed — the rendered pixels came out
  byte-identical, since nothing visible moved.)

Test plan

• npm run verify passes against python3 -m http.server 8000
• npm run design:check exits 0
• npm run og:check exits 0
• .brand .rb fill is unaffected by .brand svg{color:...} on /, /ideas/, /talks/,
  /privacy/ — the rule is added for consistency with companygraph.io/guestgraph.io, not
  because it changes anything rendered on blust.ch