Reduce required procs for some landice tests

[matthewhoffman]

This PR adjusts 3 of the landice tests (humboldt, greenland, thwaites) to require a fewer number of processors. This is done in a way that addresses 2 points of complexity:

• for decomp tests, we don't want the min_tasks logic to make both runs in the test have the same number of tasks
• we can support different numbers of procs for the FO velo solver vs. other configurations

Checklist

• User's Guide has been updated
• Developer's Guide has been updated
• API documentation in the Developer's Guide (api.rst) has any new or modified class, method and/or functions listed
• Documentation has been built locally and changes look as expected
• Document (in a comment titled Testing in this PR) any testing that was used to verify the changes

[matthewhoffman]

Testing

Prior to the commit, the landice full_integration suite had 7 test fail due to insufficient processors available when running on my Mac in the container where only 10 procs are available:

Test Runtimes:
00:14 PASS landice_dome_2000m_sia_restart_test
00:12 PASS landice_dome_2000m_sia_decomposition_test
00:05 PASS landice_dome_variable_resolution_sia_restart_test
00:03 PASS landice_dome_variable_resolution_sia_decomposition_test
00:21 PASS landice_enthalpy_benchmark_A
00:21 PASS landice_eismint2_decomposition_test
00:21 PASS landice_eismint2_enthalpy_decomposition_test
00:22 PASS landice_eismint2_restart_test
00:22 PASS landice_eismint2_enthalpy_restart_test
00:10 PASS landice_greenland_sia_restart_test
00:05 PASS landice_greenland_sia_decomposition_test
00:20 PASS landice_hydro_radial_restart_test
00:18 PASS landice_hydro_radial_decomposition_test
00:02 FAIL landice_humboldt_mesh-3km_decomposition_test_velo-none_calving-none_subglacialhydro
00:00 FAIL landice_humboldt_mesh-3km_restart_test_velo-none_calving-none_subglacialhydro
00:25 PASS landice_dome_2000m_fo_decomposition_test
00:22 PASS landice_dome_2000m_fo_restart_test
00:11 PASS landice_dome_variable_resolution_fo_decomposition_test
00:11 PASS landice_dome_variable_resolution_fo_restart_test
00:24 PASS landice_circular_shelf_decomposition_test
00:00 FAIL landice_greenland_fo_decomposition_test
02:02 PASS landice_greenland_fo_restart_test
00:00 FAIL landice_thwaites_fo_decomposition_test
01:37 PASS landice_thwaites_fo_restart_test
00:00 FAIL landice_thwaites_fo-depthInt_decomposition_test
00:46 PASS landice_thwaites_fo-depthInt_restart_test
00:00 FAIL landice_humboldt_mesh-3km_restart_test_velo-fo_calving-von_mises_stress_damage-threshold_faceMelting
00:00 FAIL landice_humboldt_mesh-3km_restart_test_velo-fo-depthInt_calving-von_mises_stress_damage-threshold_faceMelting
Total runtime 09:12
FAIL: 7 tests failed, see above.

With this branch, all tests pass:

Total runtime 12:32
PASS: All passed successfully!

In looking at how the tests were set up, I confirmed that thehumboldt/mesh-3km_decomposition_test/velo-none_calving-none_subglacialhydro decomp test is selecting the number of tasks to be the max available and half that:

10proc_run
5proc_run

[matthewhoffman]

@xylar , you don't necessarily need to make an in-depth review of this, but you might look it over and consider if there are any reasons this approach might be a concern. There is no urgency on this PR.