fix(ci): call extract_task_descriptions.py after eval in benchmark jobs

The task descriptions were never populated in metrics.json because
extract_task_descriptions.py was never invoked. The script exists and
parse_eval_metrics.py already looks for its output — the call was
simply missing from the workflow.

Appends the extraction step to the existing bash -c block (runs inside
the container where libero/metaworld is installed) so task_descriptions.json
is written to the eval-artifacts dir before docker cp copies it out.

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

.github/workflows/benchmark_tests.yml

@@ -128,6 +128,9 @@ jobs:
                 '--env.camera_name_mapping={\"agentview_image\": \"camera1\", \"robot0_eye_in_hand_image\": \"camera2\"}' \
                 --policy.empty_cameras=1 \
                 --output_dir=/tmp/eval-artifacts
+              python scripts/ci/extract_task_descriptions.py \
+                --env libero --task libero_spatial \
+                --output /tmp/eval-artifacts/task_descriptions.json
             "
 
       - name: Copy Libero artifacts from container
@@ -268,6 +271,9 @@ jobs:
                 '--rename_map={\"observation.image\": \"observation.images.camera1\"}' \
                 --policy.empty_cameras=2 \
                 --output_dir=/tmp/eval-artifacts
+              python scripts/ci/extract_task_descriptions.py \
+                --env metaworld --task metaworld-push-v3 \
+                --output /tmp/eval-artifacts/task_descriptions.json
             "
 
       - name: Copy MetaWorld artifacts from container