Always remove AWS clusters even when there are test failures

We need to identify places where we keep AWS clusters running in case of test failures.

We do this for interop pipeline, not sure if there is any other pipeline using AWS. We need to check.

Current situation is that people are not checking interop pipeline failures soon enough and the AWS cluster is running too long. We can't tolerate this as we are forced to reduce spendings on AWS.

Possible solution is to remove the cluster by default even with failed tests and add parameter to keep optionally cluster running for debugging purposes but only on demand.