Clean up E2E Operator resources if tests fail or are cancelled

In this PR, we are changing E2E Operator tests to clean up after themselves on failure or cancellation. Currently, this only happens on cancellation, but we need to do this on failure too, as tests can fail due to leaked resources.

Author: thpierce

.github/workflows/e2e-tests-with-operator.yml

@@ -189,7 +189,7 @@ jobs:
       # file locks are being held from SIGTERMS dispatched in previous
       # steps.
       - name: Destroy resources
-        if: ${{ cancelled() }}
+        if: ${{ cancelled() || failure() }}
         shell: bash {0}
         working-directory: testing-framework/terraform
         run: |
@@ -209,4 +209,4 @@ jobs:
       success: ${{  needs.run-batch-job.result == 'success' }}
       region: us-west-2
     secrets:
-      roleArn: ${{ secrets.METRICS_ROLE_ARN }}
+      roleArn: ${{ secrets.METRICS_ROLE_ARN }}