Fix test of ses(4) when there is no SES device exists

[lwhsu]

Summary

https://ci.freebsd.org/job/FreeBSD-main-amd64-test/20427/testReport/junit/sys.ses/destructive/setelmstat/
/usr/src/tests/sys/ses/destructive.c:57: r != 0

https://ci.freebsd.org/job/FreeBSD-main-amd64-test/20427/testReport/junit/sys.ses/destructive/setencstat/
/usr/src/tests/sys/ses/common.h:44: r != 0

glob(3) returns GLOB_NOMATCH if GLOB_NOCHECK or GLOB_NOMAGIC flag is not passed
so ATF_REQUIRE_EQ(r, 0) will cause a precondition check failure if no /dev/ses*
exists.

Remove calling of atf_tc_skip() in ATF_TC_CLEANUP() because it would let
the clean up procedure unfinish.

This patch calls glob(3) twice in each test case which is not optimal,
but I'd like to fix this first in CI. Maybe later we can change it to keep the
glob(3) result and then use in the rest of the test body.

Test Plan

cd /usr/tests/sys/ses && kyua test

[lwhsu]

tests/sys/ses/destructive.c

87

The original git commit message was modified by phabricator, this and the following change is to fix a set-but-not-used warning while I'm here.

[asomers]

Thanks @lwhsu ! I was actually working on this myself at this very moment. But I see that you're ahead of me.

tests/sys/ses/destructive.c

56

Should you change the flags here, too?

[lwhsu]

tests/sys/ses/destructive.c

56

This one is only called after if (!has_ses()) check failed so /dev/ses* should exist, but changing this is no harm and unify the style so I think it's fine. Do you want me also change this?

[lwhsu]

tests/sys/ses/destructive.c

56

hmm, if that's the case, then the following if (g.gl_matchc == 0) check is irrelevant, I think it's better to do defensive programming so I'll change the flag anyway.

[asomers]

tests/sys/ses/destructive.c

56

I would, just for consistency and defensiveness.

[lwhsu]

Make call to glob(3) consistency and do defensive programming