fix(cicd): use set -o pipefail instead of broken PIPESTATUS approach

PIPESTATUS[0] was consumed by the echo command before exit could use it,
causing build failures to be reported as success. set -o pipefail makes
the entire pipe fail if any command fails, which is simpler and correct.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.gitea/workflows/deploy.yaml

@@ -30,13 +30,12 @@ jobs:
       - name: Build and Push Backend
         id: build
         run: |
+          set -o pipefail
           docker buildx build \
             --push \
             --provenance=false \
             --tag ${{ secrets.SWR_SERVER }}/${{ secrets.SWR_ORG }}/rtc-backend:latest \
             . 2>&1 | tee /tmp/build.log
-          echo "build_exit_code=${PIPESTATUS[0]}" >> $GITHUB_OUTPUT
-          exit ${PIPESTATUS[0]}
 
       - name: Setup Kubectl
         run: |
@@ -70,6 +69,7 @@ jobs:
           sed -i "s|\${CI_REGISTRY_IMAGE}/backend:latest|${{ secrets.SWR_SERVER }}/${{ secrets.SWR_ORG }}/rtc-backend:latest|g" $DEPLOY_FILE
 
           # 3. 应用配置并捕获输出
+          set -o pipefail
           {
             kubectl apply -f $DEPLOY_FILE
             kubectl apply -f $INGRESS_FILE