Move Slack alert after CoreWeave teardown

Teardown should not be delayed by the Slack notification — H100 costs
matter more than a few seconds of alert latency.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: yoblin

.github/workflows/marin-canary-ferry-cw.yaml

@@ -185,15 +185,6 @@ jobs:
       # `cluster stop` only deletes Pods; NodePools survive and rely on the
       # CW autoscaler to scale down. Delete them explicitly to avoid lingering
       # H100 costs.
-      - name: Notify Slack on failure
-        if: failure() && github.event_name == 'schedule'
-        uses: slackapi/slack-github-action@v2
-        with:
-          webhook: ${{ secrets.SLACK_WEBHOOK_URL }}
-          webhook-type: incoming-webhook
-          payload: |
-            text: ":red_circle: *GPU Canary failed*\nRun: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
-
       - name: Tear down CoreWeave cluster
         if: always()
         run: |
@@ -204,3 +195,12 @@ jobs:
           else
             echo "Keeping node pool alive (keep_nodepool=true)"
           fi
+
+      - name: Notify Slack on failure
+        if: failure() && github.event_name == 'schedule'
+        uses: slackapi/slack-github-action@v2
+        with:
+          webhook: ${{ secrets.SLACK_WEBHOOK_URL }}
+          webhook-type: incoming-webhook
+          payload: |
+            text: ":red_circle: *GPU Canary failed*\nRun: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"