esmvaltool update breaks Marrmot forcing

[RolfHut]

I think an update in esmvaltool, or some other update somewhere, has broken the marrmot forcing routine in eWaterCycle.

line 73 in https://github.com/eWaterCycle/ewatercycle/blob/main/src/ewatercycle/forcing/_marrmot.py assumes that the first element in recipe_output contains the final resulting .mat file, but it does not anymore: now it points to the preproc directory.

I think line 73 needs to point more explicitly to the .mat file, without relying on the order of recipe_output to remain the constant?

See attached screenshot of the output:

[Peter9192]

This is where ewatercycle picks up the forcing file:

ewatercycle/src/ewatercycle/forcing/_marrmot.py

Lines 71 to 73 in f384c18

	# generate forcing data and retrieve useful information
	recipe_output = recipe.run(session=_session(directory))
	forcing_file: Path = list(recipe_output.values())[0].files[0].path

That index-based lookup seems a bit fragile indeed

[Peter9192]

Since ESMValGroup/ESMValCore@49c65db, preproc files are also included in the output retreived from the esmvaltool python api, unless the setting remove_preproc_dir is True.

[Peter9192]

It would be good to add a check that we indeed get a .mat file in ewatercyce/forcing/_marrmot.py