This is the first notebook that runs under the key_metric example

Removed the CESM3 vs CESM2 comparison section, refactored code to avoid
duplication.

Still to do: create utils.py so additional notebooks can share these refactored
functions. Probably some other code clean-up as well...

Created a single utils module for now

1. Reorganized cells and added markdown headers (parameter configuration, set
up grid, make datasets, generate plots).
2. Fixed issue with plotting base maps (tested by comparing first 40 years of
run to last 40 years of run)
3. Added years included in climatology to the title of plots

Commented out the atmosphere, ocean, land, and sea ice section definitions and
added a section for land ice.

Started replacing netcdf4 calls with xarray equivalents. I think the
observational dataset read is still entirely netcdf4-based, so that needs to be
updated. Also, I want to try calling xr.open_mfdataset() once per case and then
passing datasets rather than doing the read twice (once for temporal mean, once
for spatial mean).

Read in 40 years of SMB data with read_cesm_smb(), and then use xarray
functions (basically taking means or sums over different dimensions) to compute
the different climatologies. This greatly reduces the number of functions
needed in utils - it's just read_cesm_smb() and some plotting routines.

Also renamed data -> da in plot_line() to be consistent with plot_contour()

Every panel was overwriting the location of the colorbar, so only the bias
contour levels were shown. Now the first two plots (with same vmin and vmax)
have a colorbar between them while bias plot colorbar is off to the right