Fix segfault and other issues in time stepper test

[mwarusz]

This PR fixes a couple of issues in the time stepper test:

• Since this test assumes one vertical level, the config value of NVertLevels needs to be reset to 1.
• The default time stepper needs to be initialized in this test, even though it is not used. In order for it to be initialized properly the default auxiliary state needs to be initialized also.
• If two different RK4 steppers are created their provisional states need to have different names.

Unfortunately, fixing these issues does not fix the segfault on pm-gpu with gnugpu. It appears that this segfault is caused by a compiler bug. The final commit in this PR is a workaround, which fixes the segfault by removing error codes from the halo pack and unpack functions. I think this is acceptable, because there wasn't any error handling in those functions.

Checklist

• [] Testing
  • [] Unit tests have passed. Please provide a relevant CDash build entry for verification.

Aside from #183, every test is passing on pm-gpu.

[brian-oneill]

Can confirm this resolves the segfault in the TimeStepper test on pm-gpu with gnugpu. Successfully ran unit tests on Chrysalis and Frontier CPU & GPU as well (aside from #183). Other changes look good to me, too.

[mark-petersen]

Passes TIMESTEPPER_TEST on perlmutter-cpu.