Seasonal and decadal multi-models

Hi @vagudets, @nperez, @aho and @lpalma,

I have developed the first version of the multimodel for both decadal and seasonal predictions. You can find it in the branch https://earth.bsc.es/gitlab/es/auto-s2s/-/tree/dev-multimodel. I have put the recipes I used in recipes/tests/recipe_multimodel_[decadal/seasonal].yml. It has three options:

Multimodel = 'no': the different models are loaded but the multimodel is not created (dat_dim = n_models).
Multimodel = 'yes': the different models are loaded and they are combined into a multimodel (dat_dim = 1).
Multimodel = 'both': the different models are loaded, and they are combined into a multimodel. Thus, both the individual models and multimodel are kept (dat_dim = n_models+1).

Some current limitations and things to consider:

[Loading] The metadata is taken from the first model. The "important" metadata must be common among the individual models (longitudes, latitudes, dates, ...), but other metadata (e.g., loaded files) are different. I will wait for the new s2dv_cube version developed by @erifarov to be included in the tool.
[Loading] Regriding to_system is problematic if the different models have different spatial resolutions.
[Loading] The models can be initialised in months, so the lead times should be different depending on it. I would discard the first forecast months to always start in January to avoid this problem. We can talk about this in issue #9 (closed).
[Loading] Multi-seasonal means. We may want to assess the next five winters (instead of the next five years). We can also talk about this in issue #9 (closed).
[Loading] The observations are loaded while loading each model. They could only be loaded while loading the first model.
[Saving] I started working on adapting this module, but I think it is better if we first have the definitive version of the multimodel s2dv_cube.
[Visualisation] Similar to "Saving". I will work on this once we have the final multimodel s2dv_cube.

Please let me know if you have any questions.

Best regards,
Carlos