Too long test for valgrind should be leading to a failure

If a test takes more that a threshold, say 1000s, this is going to significantly impact the time that the Valgrind CI is taking. We should parse the elapsed time and make the CI fail if at least one step of takes longer than the threshold. This way, we can detect when a new test if taking too long, and avoid having Valgrind CI becoming too slow over time.