Ensure cleanup of AWS resources

[timebertt]

What this PR does / why we need it:

Before this PR, the destroy logic and testing was executed as part of a cleanup action. This is a bit dirty, as the destroy logic is part of the actual test and not just cleanup.
Also, this could lead to resource leakage on AWS and on the test-machinery cluster, because if one cleanup action fails (e.g. the destroy logic) no further cleanup actions are executed.

Now, the destroy logic is executed as a deferred function to ensure that the cleanup actions are still run, even if the test itself fails.
Additionally, a real cleanup action is added, that cleans up AWS resources left over from the test via the aws go sdk, so we can avoid manual cleanup on test failures.

Which issue(s) this PR fixes:
Fixes #

Special notes for your reviewer:
@schrodit, thanks for your help and input!

Release note:

[timebertt]

Side-note from @schrodit and me: we should refactor the infrastructure integration tests in the provider extensions in a similar way, as I already had to cleanup resources on AWS left-over from such a test manually.

[rfranzke]

/lgtm