add permissive option and fix for 0

[juliettelavoie]

Pull Request Checklist:

• This PR addresses an already opened issue (for bug fixes / features)
  • This PR fixes #xyz
• (If applicable) Documentation has been added / updated (for bug fixes / features).
• (If applicable) Tests have been added.
• CHANGELOG.rst has been updated (with summary of main changes).
  • Link to issue (:issue:number) and pull request (:pull:number) has been added.

What kind of change does this PR introduce?

• Fix clip for 0. When we had a 0 in float32 and we clipped to the next float32, it would become a 0 (float 32) again in data_prime. If we cast it to float64, it could stay not exactly 0.
  • our tests did not catch this because the inputs were in 64.
  • The casting behavior is different for numpy <2.0, so I pin it above.
• I have encountered climate models with hurs above 100... I have found that the data_request suggested to clip the data, but not everybody did it. (https://earthscience.stackexchange.com/questions/22850/why-are-the-cmip6-model-outputs-for-relative-humidity-not-scaled-to-between-1-10)
  • I am thinking maybe we can do the clipping ourselves, so I added that option with "permissive".

Does this PR introduce a breaking change?

yes, the behavior will be different for 0.

Other information:

[coveralls]

coverage: 78.173%. remained the same
when pulling 0cd5f1e on fix-clip0
into 251cd9c on main.

[juliettelavoie]

The "permissive" option was necessary when clipping before the logit. I don't think we should clip anymore so this is kind of irrelevant now. But, I am waiting for people (genre Pierre et Martin) to approve the document on hurs before giving up completely on it.

In that case, I would still merge but only with the Fix clip for 0 part. so stand-by for now.