ci: run benchmark job after test jobs to avoid GPU contention

Benchmark was running in parallel with all 4 test jobs, putting 5
concurrent GPU workloads on the same runner and causing OOM for
everything. Sequencing benchmark after test ensures it gets the
GPU to itself.

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

Author: camille-004

.github/workflows/ci.yml

@@ -191,7 +191,7 @@ jobs:
 
   benchmark:
     name: Benchmark
-    needs: [images, build]
+    needs: [images, build, test]
     runs-on: [self-hosted, gpu]
     timeout-minutes: 60
     continue-on-error: true