ci(deploy): cache-bust smoke test against Cloudflare edge

The post-deploy smoke test was fetching www.cameleer.io/ directly,
which Cloudflare may serve from edge cache regardless of what rsync
just shipped — falsely greenlighting a failed deploy or red-flagging
a successful one. Append ?cb=$GITHUB_RUN_ID and Cache-Control:
no-cache to both fetches so the edge revalidates per run.

Surfaced by code review of the prior commit (4c98caab).

.gitea/workflows/deploy-placeholder.yml

@@ -89,11 +89,15 @@ jobs:
         run: |
           set -e
           echo "Confirming the placeholder is live on www.cameleer.io..."
-          BODY=$(curl -sf https://www.cameleer.io/)
+          # Cache-bust per run so Cloudflare's edge can't serve a stale response
+          # that masks a failed deploy. ?cb=$GITHUB_RUN_ID forces a fresh cache key;
+          # the no-cache request header tells any well-behaved cache to revalidate.
+          CB="$GITHUB_RUN_ID"
+          BODY=$(curl -sf -H 'Cache-Control: no-cache' "https://www.cameleer.io/?cb=$CB")
           echo "$BODY" | grep -qF 'Routes are remapping' \
             || { echo "Sentinel string missing — placeholder did not land."; exit 1; }
           echo "$BODY" | grep -qF 'mailto:' \
             || { echo "mailto: link missing — sales email substitution may have failed."; exit 1; }
-          curl -sfI https://www.cameleer.io/cameleer-logo.png > /dev/null \
+          curl -sfI -H 'Cache-Control: no-cache' "https://www.cameleer.io/cameleer-logo.png?cb=$CB" > /dev/null \
             || { echo "cameleer-logo.png not reachable on the live origin."; exit 1; }
           echo "Placeholder is live."