Optimize failure() condition for pre/post steps

Pre/post steps support condition `failure()` which runs the step only on a failed test.

When running a test group with multiple tests the failure() condition is evaluated on the whole test suite. This leads to the situation that the `failure()` condition evaluates to `true` for the whole test suite once a single test has failed. So all succeeding tests will run the steps using the failed condition regardless of the individual test result (success/failed).

We should change that behavior so the pre/post steps that are bound to failure state are run only for failed tests. Succeeding tests without failure should not run the steps in their pre/post script section because of this.