fix(ci): 배포 시 jar 경로 자가치유 (컨테이너 stop→경로 정리→배치→재기동)

빈 디렉토리로 잘못 생성된 마운트 경로(corrupt jarfile 원인)까지 rm -rf 로 정리.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.gitea/workflows/deploy-prod.yaml

@@ -58,7 +58,9 @@ jobs:
 
           $SSH "$DEPLOY_USER@$DEPLOY_HOST" "mkdir -p \"\$(dirname '$DEPLOY_JAR_PATH')\" && cat > '$DEPLOY_JAR_PATH.new'" < "$JAR"
           $SSH "$DEPLOY_USER@$DEPLOY_HOST" "\
+            docker compose -f '$COMPOSE_FILE' stop api 2>/dev/null || true; \
+            rm -rf '$DEPLOY_JAR_PATH'; \
             mv -f '$DEPLOY_JAR_PATH.new' '$DEPLOY_JAR_PATH'; \
             docker compose -f '$COMPOSE_FILE' up -d --force-recreate api; \
-            docker compose -f '$COMPOSE_FILE' ps api"
+            sleep 2; docker compose -f '$COMPOSE_FILE' ps api"
           echo "✅ 운영(api) 재기동 완료"