Use linux.8xlarge.memory runner for Cadence tests

Larger runner to avoid OOM kills from parallel cadence_runner
subprocesses, and restore -n auto for test-ops.

Author: aliafzal

.github/workflows/_test_cadence.yml

@@ -12,7 +12,7 @@ on:
         description: 'Runner type'
         required: false
         type: string
-        default: linux.2xlarge.memory
+        default: linux.8xlarge.memory
       ref:
         description: 'Git ref to checkout'
         required: false
@@ -70,4 +70,4 @@ jobs:
         chmod +x cmake-out/backends/cadence/cadence_runner
 
         export PYTHONPATH="${PYTHONPATH:-}:$(pwd)/backends/cadence/utils/FACTO"
-        python -m pytest examples/cadence/operators/ -v -n 4
+        python -m pytest examples/cadence/operators/ -v -n auto