Add rollback procedure documentation and testing scripts (FRE-4808)

- infra/ROLLBACK.md: comprehensive rollback runbook with ECS, Docker Compose,
  database migration, blue-green, and emergency rollback procedures
- infra/scripts/rollback.sh: enhanced ECS rollback with validation, logging,
  health verification, and per-service rollback support
- infra/scripts/rollback-compose.sh: Docker Compose rollback for local/staging
- infra/scripts/rollback-migration.sh: Drizzle migration rollback with
  AWS Secrets Manager integration
- infra/scripts/test-rollback.sh: automated test suite (51 tests)
- Updated infra/README.md to reference ROLLBACK.md

Co-Authored-By: Paperclip <noreply@paperclip.ing>

infra/scripts/rollback-compose.sh

@@ -0,0 +1,121 @@
+#!/bin/bash
+set -euo pipefail
+
+# ShieldAI Docker Compose Rollback Script
+# Usage: ./rollback-compose.sh <previous_tag> [--env prod|dev]
+#
+# Rolls back all services to a previous tagged image using docker-compose.prod.yml
+#
+# Examples:
+#   ./rollback-compose.sh v1.2.3              # Rollback to v1.2.3
+#   ./rollback-compose.sh v1.2.3 --env prod   # Explicit production compose
+
+PREVIOUS_TAG="${1:-}"
+ENV_MODE="${2:-prod}"
+
+# ─── Configuration ───────────────────────────────────────────────
+SERVICES="api darkwatch spamshield voiceprint"
+COMPOSE_FILE="docker-compose.prod.yml"
+REGISTRY_OWNER="${GITHUB_REPOSITORY_OWNER:-shieldai}"
+
+# ─── Helpers ─────────────────────────────────────────────────────
+log() {
+  local level="$1"
+  shift
+  echo "[$(date -u '+%H:%M:%S')] [$level] $*"
+}
+
+log_info()  { log "INFO" "$@"; }
+log_warn()  { log "WARN" "$@"; }
+log_error() { log "ERROR" "$@"; }
+
+# ─── Validation ──────────────────────────────────────────────────
+if [[ -z "$PREVIOUS_TAG" ]]; then
+  log_error "Usage: $0 <previous_tag> [--env prod|dev]"
+  log_error "Example: $0 v1.2.3"
+  exit 1
+fi
+
+if ! command -v docker &>/dev/null; then
+  log_error "Docker not found in PATH"
+  exit 1
+fi
+
+# ─── Rollback Logic ──────────────────────────────────────────────
+main() {
+  log_info "=== Docker Compose Rollback ==="
+  log_info "Target tag: $PREVIOUS_TAG"
+  log_info "Compose file: $COMPOSE_FILE"
+  log_info "Registry: ghcr.io/$REGISTRY_OWNER"
+
+  # 1. Pull previous images
+  log_info "Pulling previous images..."
+  local pull_failed=0
+  for svc in $SERVICES; do
+    local image="ghcr.io/${REGISTRY_OWNER}/shieldai-${svc}:${PREVIOUS_TAG}"
+    log_info "Pulling $image..."
+    if docker pull "$image" 2>/dev/null; then
+      log_info "Pulled: $image"
+    else
+      log_warn "Pull failed: $image (may not exist)"
+      pull_failed=1
+    fi
+  done
+
+  if [[ $pull_failed -eq 1 ]]; then
+    log_warn "Some images may not exist at tag $PREVIOUS_TAG"
+    log_info "Continuing with available images..."
+  fi
+
+  # 2. Stop current services gracefully
+  log_info "Stopping current services..."
+  DOCKER_TAG="$PREVIOUS_TAG" docker compose -f "$COMPOSE_FILE" down --timeout 30 2>/dev/null || true
+
+  # 3. Start with previous tag
+  log_info "Starting services with tag $PREVIOUS_TAG..."
+  DOCKER_TAG="$PREVIOUS_TAG" docker compose -f "$COMPOSE_FILE" up -d
+
+  # 4. Wait for services to be healthy
+  log_info "Waiting for services to become healthy..."
+  sleep 10
+
+  # 5. Verify health
+  local passed=0
+  local failed=0
+
+  for svc in $SERVICES; do
+    local port
+    port=$(case "$svc" in
+      api) echo 3000 ;;
+      darkwatch) echo 3001 ;;
+      spamshield) echo 3002 ;;
+      voiceprint) echo 3003 ;;
+    esac)
+
+    local http_code
+    http_code=$(curl -s -o /dev/null -w "%{http_code}" \
+      --connect-timeout 10 --max-time 30 \
+      "http://localhost:${port}/health" 2>/dev/null || echo "000")
+
+    if [[ "$http_code" == "200" ]]; then
+      log_info "Health OK: $svc (port $port, HTTP $http_code)"
+      ((passed++))
+    else
+      log_warn "Health FAIL: $svc (port $port, HTTP $http_code)"
+      ((failed++))
+    fi
+  done
+
+  log_info "=== Rollback Complete ==="
+  log_info "Passed: $passed, Failed: $failed"
+
+  if [[ $failed -gt 0 ]]; then
+    log_warn "Some services failed health check. Check logs: docker compose -f $COMPOSE_FILE logs"
+    exit 1
+  fi
+
+  log_info "All services healthy after rollback"
+  exit 0
+}
+
+main "$@"