Remove lmoments3

[aulemahal]

Pull Request Checklist:

• This PR addresses an already opened issue (for bug fixes / features)
  • This PR fixes Integration of lmoments3 and eofs is not permitted under GPLv3 #1620
• Tests for the changes have been added (for bug fixes / features)
  • (If applicable) Documentation has been added / updated (for bug fixes / features)
• CHANGES.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?

• Remove lmoments3 from xclim's dependencies.
• Accept rv_continuous instances for the dist arguments of the statistical indices

Does this PR introduce a breaking change?

Yes. Passing a mere string and method='PWM' is now broken. It will raise a ValueError with a message asking to pass an instance of lmoments3 instead.

This also makes some functions awkward to use. Before, we relied on the scipy_dist attribute of params to retrieve the distribution when computing statistics. As one can now pass an object that has nothing to do with scipy, those functions (parametric_quantile for example) must now also accept the dist as an argument.

Other information:

See Ouranosinc/lmoments3#12.

I kinda cheated and made it so the dist argument would show up as a "String" parameter to the indicator. This only changes the metadata, distributions objects can still be passed to indicators.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[github-actions]

Note
It appears that this Pull Request modifies the main.yml workflow.

On inspection, the XCLIM_TESTDATA_BRANCH environment variable is set to the most recent tag (v2023.12.14).

No further action is required.

[huard]

Looks good. Could you confirm stats.frequency_analysis still works ?

[aulemahal]

@huard
This code works:

import xclim as xc
import xarray as xr
import numpy as np
import lmoments3.distr as lmom

q = xr.DataArray(
    np.random.random_sample((365 * 10 + 2,)), 
    dims=('time',),
    coords={'time': xr.cftime_range('1990-01-01', freq='D', periods=365 * 10 + 2)},
    attrs={'units': 'm3 s-1'}
)

xc.generic.return_level(q, mode='max', t=2, dist=lmom.gev, window=6, freq='YS', method='PWM')
xc.generic.return_level(q, mode='max', t=2, dist='genextreme', window=6, freq='YS', method='ML')

[aulemahal]

@saschahofmann If you did not receive news from the last person, does this PR seems reasonable to you ?

[coveralls]

coverage: 90.164% (+0.004%) from 90.16%
when pulling 4a51936 on remove-lmoments
into d09bd4b on master.

[saschahofmann]

Yes totally . Tomorrow I'll be a little rude and give him a nudge on the lmoments R library that he still maintains hoping we can get his attention that way

[Zeitsperre]

Nicely done!