[query] Run benchmarks after wheel is published to PyPI (#14766)

We publish release artifacts independently of tests passing in the
release pipeline. Flaky tests were causing the benchmarks to not run.
This change takes the pragmatic view that if publishing does not depend
on tests, then neither should starting the benchmark batch.

## Security Assessment
- This change has no security impact

build.yaml

@@ -3735,13 +3735,6 @@ steps:
     dependsOn:
       - default_ns
       - hail_dev_image
-      - test_hail_java
-      - test_hail_python_local_backend
-      - test_hail_python_service_backend_gcp
-      - test_hail_python_service_backend_azure
-      - test_hail_scala_fs
-      - test_hail_services_java
-      - test_hail_spark_conf_requester_pays_parsing
       - release
     image:
       valueFrom: hail_dev_image.image