Climetlab-s2s-ai-challenge

- observations are now regridded with `xesmf` `method='conservative'`: https://renkulab.io/gitlab/aaron.spring/s2s-ai-challenge/-/issues/32 PLEASE DOWNLOAD OBSERVATIONS AGAIN. THE FORMER DATA WILL NOT BE USED IN THE S2S-AI-CHALLENGE.
- observations with a dimension are now additionally available via `grid='720x360'`: https://renkulab.io/gitlab/aaron.spring/s2s-ai-challenge/-/issues/32
- fixed all `NaN` longitude=0 in observations: https://renkulab.io/gitlab/aaron.spring/s2s-ai-challenge/-/issues/35
- fixed all `NaN` maps after leap day: https://renkulab.io/gitlab/aaron.spring/s2s-ai-challenge/-/issues/31

- create `tp` and `t2m` `hindcast-like-observations`/`forecast-like-observations` locally from `pr` and `t2m` `observations` with `climetlab_s2s_ai_challenge.extra.forecast_like_observations`
- data uploaded to EWC changed: `tp` variable from `training-output-reference` and `test-output-reference` now calculated with `climetlab_s2s_ai_challenge.extra.forecast_like_observations`. lead_time is now shifted one day. If you downloaded these datasets before, please [clear/delete your cache folder](https://climetlab.readthedocs.io/en/latest/developer/caching.html) and download new, see [climetlab settings](https://climetlab.readthedocs.io/en/latest/guide/settings.html) for how to manage cache and get threaded downloads.
- restructuring of the README

Fixed values of lead_time in the benchmark datasets.