Fix mom-tx10deg/tx3deg testmods for non-tx2_3v2 grids (#338)

[alperaltuntas]

Fixes: #338.

The mom-tx10deg and mom-tx3deg testmods failed at runtime on any ocean grid other than tx2_3v2 such as tx2_3v3 because GRID_FILE, TOPO_FILE, CHL_FILE and SALT_RESTORE_FILE were given as paths relative to INPUTDIR which is DIN_LOC_ROOT ocn mom OCN_GRID and the coarse resolution files only exist under the tx2_3v2 input directory, so MOM6 looked under a nonexistent directory and aborted. This was never caught in CI because every test using these testmods runs on a t232 grid.

This change points the four files at the tx2_3v2 directory using a relative path so they resolve correctly regardless of OCN_GRID, and a relative path is used rather than DIN_LOC_ROOT because MOM_override does no variable expansion. Mesh and CICE files were already hardcoded to tx2_3v2 via shell_commands and are unchanged. Verified the resolved paths exist for both tx2_3v3 and tx2_3v2.

[mnlevy1981]

Should the tx3deg and tx10deg directories be in grid_indpt/ instead of tx2_3v2/?

[alperaltuntas]

Should the tx3deg and tx10deg directories be in grid_indpt/ instead of tx2_3v2/?

The reason they are under tx2_3v2/ is because they are subsamplings of tx2_3v2/ by factors of 5 and 15 respectively. However, If you feel strongly about moving them to grid_indpt/`, I can do so.

[mnlevy1981]

Should the tx3deg and tx10deg directories be in grid_indpt/ instead of tx2_3v2/?

The reason they are under tx2_3v2/ is because they are subsamplings of tx2_3v2/ by factors of 5 and 15 respectively. However, If you feel strongly about moving them to grid_indpt/`, I can do so.

I don't have strong feelings, it just seemed like grid_indpt was the right place for something that should work for any ocean grid. Given its relationship to the 2/3° grid, the current location makes sense

[kdraeder]

Mesh and CICE files were already hardcoded to tx2_3v2 via shell_commands and are unchanged.

Will this be the best solution for the foreseeable future?
Or is this a good time to move their definitions from shell_commands to these xlm files?

Other than that, your testing shows that these changes solve the problem, so I'll approve them.

[alperaltuntas]

Mesh and CICE files were already hardcoded to tx2_3v2 via shell_commands and are unchanged.

Will this be the best solution for the foreseeable future? Or is this a good time to move their definitions from shell_commands to these xlm files?

Other than that, your testing shows that these changes solve the problem, so I'll approve them.

These are ad-hoc testmods that aren't supported or endorsed for regular use, so I'd caution against moving them into the xml files, which could give the false impression that they're officially supported.

[kdraeder]

Thanks for fixing this!