Add Large PE layouts for mpasa3p75 testing

[ekluzek]

Description of changes

Add two new PE layouts for mpas3p75 a "L" and "Lhm" option.
Add tests of these to the uhr_decomp_init test list

Specific notes

Contributors other than yourself, if any: @briandobbins

CTSM Issues Fixed (include github issue #):
Fixes #3489

Are answers expected to change (and if so in what way)? No

Any User Interface Changes (namelist or namelist defaults changes)? Yes

Allows using the "L" or "Lhm" option for the PE layout for the mpasa3p75 grid

Does this create a need to change or add documentation? Did you do so? No No

Testing performed, if any: So far just running the uhr_decomp_init test list

[ekluzek]

Here's the init results for these two, which clearly show the excessive amount of time being spend in decompInit. This is based off of b4b-dev so doesn't have the extra timers, but shows the principle. It also shows how having more memory on each node saves a significant amount of time.

For mpasa3p75 L:

  [ESMF]                                                                     38400  38400  1        2261.4802   2261.4680   30547   2261.4900   30516
    [ensemble] Init 1                                                        38400  38400  1        2242.6065   2242.4292   35762   2242.7544   11944
      [ESM0001] IPDv02p3                                                     38400  38400  1        2183.6020   2183.5979   27186   2183.6047   26932
        [LND] IPDv01p3                                                       38400  38400  1        2001.9617   2001.9426   18534   2001.9717   37135
          lc_lnd_set_decomp_and_domain_from_readmesh                         38400  38400  1        1971.5981   1971.5962   7759    1971.6025   4342

For mpasa3p75 Lhm:

  [ESMF]                                                                     38400  38400  1        1102.8676   1102.8573   35995   1102.8741   34724
    [ensemble] Init 1                                                        38400  38400  1        1085.9691   1085.8829   22560   1086.0488   25615
      [ESM0001] IPDv02p3                                                     38400  38400  1        1039.9769   1039.9738   31175   1039.9792   36776
        [LND] IPDv01p3                                                       38400  38400  1        900.3199    900.3007    3036    900.3278    20220
          lc_lnd_set_decomp_and_domain_from_readmesh                         38400  38400  1        873.0665    873.0646    5301    873.0708    9880

I'll get this on a branch with the extra timers so we can see the details of how this performs against the new case with mpi_scan. Based on our other results the rest of lc_lnd_set_decomp_and_domain_from_readmesh should be around 20 seconds. So even if you quadruple that you have an enormous amount of time spent in decompInit_lnd.

[slevis-lmwg]

@ekluzek I like the comments, but I don't like the use of Clm45 compsets in the new tests. I'm approving, and you can decide what's best.