about 10-20% of times smoketests were using this script to ensure availability
of services, the smoketest failed after the script succeeded. This change
improves the script by:

* providing an early exit for already deployed services without testing
  multiple times for success if the first test succeeds
* raising the amount of successful checks needed to 25 and sleeping a second
  after each successful test in order to lower the probability of the service
  not being available for the smoketest after the script succeeds.

terraform destroy failed on the first run, because it tried to delete
the db user before the database, which still had dependencies on the
user

otherwise terraform always wants to apply changes to the function