2D ensemble weights

[RondeauG]

Pull Request Checklist:

• [x] This PR addresses an already opened issue (for bug fixes / features)
  • This PR fixes #108
  • This PR fixes #230
• [x] (If applicable) Documentation has been added / updated (for bug fixes / features).
• [x] (If applicable) Tests have been added.
• [x] This PR does not seem to break the templates.
• [x] HISTORY.rst has been updated (with summary of main changes).
  • [x] Link to issue (:issue:number) and pull request (:pull:number) has been added.

What kind of change does this PR introduce?

• Major changes to generate_weights:
  • The function itself was mostly rewritten, in an effort to more easily support independence levels.
  • The results for independence_level: GCM were changed substantially. See #230.
  • New independence level: institution
  • New argument: split_experiments: Pretty self-explanatory
  • New argument: skipna:
  • If True (default), then the weights are built using the attributes only.
  • If False, the weights are built by taking into account the values themselves and putting a weight of 0 when NaNs are encountered. Requires a time or horizon dimension. If the dimension is not the same across all datasets, they are first reindexed to cover all horizon/time.
• Updated produce_horizon so it can accept multiple periods or warming levels.
• Small bugfix in cleanup to prevent raising an error when no cat: attribute exists in the datasets.
• Adds tests to most of xs.ensemble_stats.
• Adds tests to xs.generate_weights
• Updated the notebooks to reflect the changes (Getting Started - Ensemble stats & Warming Levels)

Does this PR introduce a breaking change?

• Removed a FutureWarning in ensemble_stats that was there since xscen 0.4.0. stats_kwargs is not fully abandoned.
• independence_level: all was renamed model for better consistency.
• The results for independence_level: GCM were changed substantially. See #230.
• period in produce_horizon has been deprecated and replaced with periods.
• Some automated to_level were updated to reflect more recent changes.

Other information:

[RondeauG]

I can't turn the PR into a draft, but you can wait a little before reviewing this. While updating the Notebooks, I realized that produce_horizon could be updated to allow multiple periods/warming levels, now that we can support that further down the road.

[RondeauG]

Alright, this is good for review.

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB