test: wait a bit after starting the test service

otherwise we might end up being faster than the unit itself, causing
unexpected fails, like:

```
testsuite.sh[297]: + systemctl start issue_14566_test
testsuite.sh[297]: + systemctl status issue_14566_test
testsuite.sh[304]: ● issue_14566_test.service - Issue 14566 Repro
testsuite.sh[304]:      Loaded: loaded (/etc/systemd/system/issue_14566_test.service; static; vendor preset: enabled)
testsuite.sh[304]:      Active: active (running) since Sat 2020-03-14 02:02:23 UTC; 417ms ago
testsuite.sh[304]:    Main PID: 301 ((repro.sh))
testsuite.sh[304]:       Tasks: 1 (limit: 535)
testsuite.sh[304]:      Memory: 180.0K
testsuite.sh[304]:         CPU: 122ms
testsuite.sh[304]:      CGroup: /system.slice/issue_14566_test.service
testsuite.sh[304]:              └─301 [(repro.sh)]
testsuite.sh[307]: ++ cat /leakedtestpid
testsuite.sh[307]: cat: /leakedtestpid: No such file or directory
testsuite.sh[297]: + leaked_pid=
```

(cherry picked from commit 197298ff)