Markets: every AP3X venue, live, with the proper marks

The Markets section listed three Base pairs from DexScreener and an
empty "Prime tokens" box. It now shows every venue where AP3X trades:

- Base: Aerodrome USDC (the reference price), QuickSwap cbBTC, the two
  Uniswap V3 pools DexScreener does not index and the Aerodrome cbADA
  pool it lists without a price - the last three read straight from
  the chain in one batched eth_call (publicnode, CORS-open; the
  out-of-range WETH pool is flagged and earns no TVL credit)
- Cardano: VyFi ADA/cstAP3X, ADA/cAP3X, cAP3X/cstAP3X from Koios
  address_info (reserves summed across UTxOs), Minswap V2 cAP3X/ADA and
  NIGHT/cAP3X from the pool datum found by the LP asset name, priced by
  the Minswap aggregator; DexHunter stands in when Koios is down
- LBank AP3X/USDT through the site-origin proxy, shown as what it is:
  no trades in days, spread, dead volume
- USDC.e on Nexus gets its own card (ApexScan: supply, holders, largest
  holder, all-time transfers, Stargate bridge link) in place of the
  Prime tokens box, which had nothing to list
- "One asset, four chains": AP3X, bAP3X, cAP3X, cstAP3X with the
  cstAP3X redemption rate computed live (vault / issued) so the
  receipt's premium is explained, not mispriced
- marks are the HAL 8 wallet's own PNGs plus each venue's mark, token
  image with a chain badge, embedded so the page stays one file
- feeds have deadlines, per-panel parsers, an error boundary, tiered
  refresh (quota'd feeds every 9 min, fast ones every 90 s), hidden-tab
  pause, a refresh cooldown, and honest fallbacks (ADA/USD from Minswap
  when CoinGecko is down, cAP3X/USD from reserves when the aggregator
  429s, placeholders when DexScreener is down)
- chain pulse now says mainnet; HeroLightning's ResizeObserver no longer
  throws after unmount; stake-pool saturation reads OSAM's field

Koios and LBank are reached via /api/koios/* and /api/lbank/* on this
origin (Caddy reverse proxies in the HAL 8 repo, PR #142) because
neither answers with a usable CORS header.

README documents the feeds; tools/bundle.py decodes and re-encodes the
bundled page byte-exactly so the next edit does not touch escaped text.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UPpFgu6wXiDfJv8ghoUwBJ

tools/bundle.py

@@ -0,0 +1,42 @@
+#!/usr/bin/env python3
+"""Decode / re-encode the __bundler/template of the apex-fusion site.
+
+  python3 bundle.py decode  index.html  page.html     # template -> editable HTML
+  python3 bundle.py encode  index.html  page.html  out.html   # editable HTML -> site file
+
+The bundler stores the page as a JSON string with '</' written as '<\\u002F'
+(so the inner page's </script> tags cannot terminate the outer tag). Encoding
+mirrors that exactly, and `encode` proves it by asserting that an untouched
+page round-trips byte-identical before writing.
+"""
+import json, re, sys
+
+TAG = re.compile(r'(<script type="__bundler/template">)(.*?)(</script>)', re.S)
+
+def decode(site):
+    return json.loads(TAG.search(site).group(2))
+
+def encode_template(page):
+    e = json.dumps(page, ensure_ascii=False)
+    return '\n' + e.replace('</', '<\\u002F') + '\n  '
+
+def encode(site, page):
+    m = TAG.search(site)
+    return site[:m.start(2)] + encode_template(page) + site[m.end(2):]
+
+if __name__ == '__main__':
+    cmd, site_path, page_path = sys.argv[1:4]
+    site = open(site_path, encoding='utf-8').read()
+    if cmd == 'decode':
+        open(page_path, 'w', encoding='utf-8').write(decode(site))
+        print('decoded', len(decode(site)), 'chars')
+    elif cmd == 'encode':
+        out_path = sys.argv[4]
+        # fidelity proof: the original template must re-encode to itself
+        orig_block = TAG.search(site).group(2)
+        assert encode_template(decode(site)) == orig_block, 'encoder no longer byte-exact for the untouched page'
+        page = open(page_path, encoding='utf-8').read()
+        out = encode(site, page)
+        assert decode(out) == page, 'round-trip mismatch after encode'
+        open(out_path, 'w', encoding='utf-8').write(out)
+        print('encoded', len(out), 'bytes ->', out_path)