Recipes fail when using matplotlib v3.9.0

[ehogan]

As a final sanity check, I reran all the tests overnight last night using ESMValCore v2.11.0. I'm glad I did! A number of recipes failed due to a newly-released version of matplotlib 😭

(old-env) % conda list matplotlib
# packages in environment at /miniforge3/envs/old-env:
#
# Name                    Version                   Build  Channel
matplotlib-base           3.8.4           py311ha4ca890_2    conda-forge
(new-env) % conda list matplotlib
# packages in environment at /miniforge3/envs/new-env:
#
# Name                    Version                   Build  Channel
matplotlib-base           3.9.0           py311hffb96ce_0    conda-forge

• AttributeError: module 'matplotlib.cm' has no attribute 'get_cmap':

  • recipe_aod_aeronet_assess.yml
  • recipe_autoassess_stratosphere.yml
  • recipe_esacci_oc.yml
  • recipe_monitor.yml
  • recipe_ocean_Landschuetzer2016.yml
  • recipe_ocean_bgc.yml
  • recipe_ocean_example.yml
  • recipe_ocean_multimap.yml
  • recipe_ocean_quadmap.yml
  • recipe_ocean_scalar_fields.yml
  • recipe_psyplot.yml

• AttributeError: 'Legend' object has no attribute 'legendHandles'. Did you mean: 'legend_handles'?

  • recipe_mpqb_xch4.yml

• AttributeError: 'GeoContourSet' object has no attribute 'labelTextsList'. Did you mean: 'labelLevelList'?

  • recipe_zmnam.yml

I would like to propose, to avoid any further testing (and given this is my last day working on the release!), that we pin matplotlib to <3.9.0 on the ESMValTool release branch, but that we leave this issue open to resolve the issue more permanantly (similarly to #3701) on main.