Fix how yaml model config options are update

[xylar]

With this update, we treat config options from files (with sections) differently from replacement options provided in a dictionary in code.

This merge also explicitly converts some model config options to floats. This seems to be necessary because the yaml parser doesn't understand numpy scalars.

Checklist

• Testing comment in the PR documents testing used to verify the changes

[xylar]

Testing

With these instructions, I was able to set up the manufactured solution test case for Omega following @hyungyukang's instructions modified here for Intel on Chrysalis:

workdir=$PWD
git clone -b hkang/omega-sw-try1 https://github.com/hyungyukang/Omega.git
cd Omega
git submodule update --init --recursive
cd components/omega/build

module load cmake

PARMETIS_ROOT=/lcrc/soft/climate/polaris/chrysalis/spack/dev_polaris_0_4_0_intel_openmpi/var/spack/environments/dev_polaris_0_4_0_intel_openmpi/.spack-env/view
cmake \
   -DOMEGA_CIME_COMPILER=intel  \
   -DOMEGA_CIME_MACHINE=chrysalis \
   -DOMEGA_PARMETIS_ROOT=${PARMETIS_ROOT}\
   -Wno-dev \
   ../

./omega_build.sh

cp -rf ./default_inputs ./src/

cd $workdir
git clone -b fix-model-yaml-update git@github.com:xylar/polaris.git
cd polaris

./configure_polaris_envs.py --conda ~/miniforge3 -c intel -m crhysalis

cp ../Omega/components/omega/build/default_inputs/omega.cfg polaris/ocean/

vim polaris/ocean/ocean.cfg

# Change line 7 from 'mpas-ocean' to 'omega'
# model = omega

source load_dev_polaris_0.4.0-alpha.2_chrysalis_intel_openmpi.sh
polaris setup -t ocean/planar/manufactured_solution -m pm-cpu -w /lcrc/group/e3sm/ac.xylar/polaris_0.4/chrysalis/test_20240807/omega_manufactured_solution -p ../Omega/components/omega/build/src

[xylar]

@hyungyukang, could you let me know if this fixes the problem for you as well?

[hyungyukang]

@xylar , thank you very much for your quick update! This PR fixes the problem I had on pm-cpu.

After setting up the manufactured solution test with the Omega-0 test model, I submitted a job and was able to get this convergence result:

Note that the model options are hardcoded in the test model now, and time step size is fixed.

I'm approving this PR as the yaml generation problem has been fixed.

Thanks again for your work on this!

[xylar]

@hyungyukang, thanks for the quick review! Glad it worked for you, too!