ops(docker): prune unused cache on deploys and nightly

Add scripts/docker-prune.sh (build cache >72h capped at 4g, unreferenced images >7d, stopped containers >24h; never volumes), run it after every CI deploy and compose update, and schedule a nightly prune from the host-side jobs-worker. Tighten the deployment contract tests to assert the scoped-prune boundaries.

scripts/docker-prune.sh

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+# Reclaim Docker's unused cache so host storage stays bounded.
+#
+# Safe scopes only, by design:
+#   - BuildKit cache older than 72h, hard-capped at 4 GB (Debian /pnpm store is
+#     shared across builds; everything newer than that speeds up rebuilds).
+#   - Images referenced by NO running/stopped container and older than 7 days
+#     (covers stale epicnext-cms sha tags, old mariadb/byparr pulls, etc.).
+#   - Containers stopped for more than 24h.
+#
+# Volumes are NEVER pruned here: mariadb-turbo-data is a database. This script
+# is idempotent and exits 0 when Docker is unavailable.
+set -Eeuo pipefail
+DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+LOG_DIR="${LOG_DIR:-$DIR/logs}"
+mkdir -p "$LOG_DIR"
+LOG_FILE="$LOG_DIR/docker-prune.log"
+now() { date '+%Y-%m-%d %H:%M:%S'; }
+
+command -v docker >/dev/null 2>&1 || {
+  printf '[%s] docker CLI unavailable; nothing to prune\n' "$(now)" >>"$LOG_FILE"
+  exit 0
+}
+
+printf '\n[%s] === docker prune start ===\n' "$(now)" >>"$LOG_FILE"
+docker system df >>"$LOG_FILE" 2>&1 || true
+
+docker builder prune -af --filter "until=72h" --max-used-space=4g >>"$LOG_FILE" 2>&1 || true
+docker image prune -af --filter "until=168h" >>"$LOG_FILE" 2>&1 || true
+docker container prune -f --filter "until=24h" >>"$LOG_FILE" 2>&1 || true
+
+printf '\n[%s] === docker prune complete ===\n' "$(now)" >>"$LOG_FILE"
+docker system df >>"$LOG_FILE" 2>&1 || true