fix: health check error handling again

[Maks1mS]

The supervisorctl status command may return status 3, which causes the process to terminate prematurely due to set -e (228558f).

weblate@05c0208c750d:/$ services="$(/app/venv/bin/supervisorctl status)"
weblate@05c0208c750d:/$ echo $?
3
weblate@05c0208c750d:/$ echo $services
celery RUNNING pid 337, uptime 0:13:17 celery-beat RUNNING pid 338, uptime 0:13:17 check EXITED Dec 21 10:22 AM granian RUNNING pid 335, uptime 0:13:17 nginx RUNNING pid 336, uptime 0:13:17

$ sh -c 'echo start; f() { return 3; }; f; echo end;'
start
end
$ sh -c 'set -e; echo start; f() { return 3; }; f; echo end;'
start

Possible fix:

weblate@05c0208c750d:/$ services="$(/app/venv/bin/supervisorctl status || true)"
weblate@05c0208c750d:/$ echo $?
0
weblate@05c0208c750d:/$ echo $services
celery RUNNING pid 337, uptime 0:14:00 celery-beat RUNNING pid 338, uptime 0:14:00 check EXITED Dec 21 10:22 AM granian RUNNING pid 335, uptime 0:14:00 nginx RUNNING pid 336, uptime 0:14:00

[nijel]

This was introduced because supervisorctl was broken. By ignoring the exit status completely it will again be unnoticed

[nijel]

#3999 should be a better approach; review is welcome!