add warning for empty bins in ClusterEnsemble

[m-aguena]

Closes issue #618 , second interaction

[coveralls]

coverage: 100.0%. remained the same
when pulling fa84ad6 on issue/618/empty_radial_bin_again
into cfa7f11 on main.

[marina-ricci]

Just to copy here what we discussed : we want to return NaN instead of 0 when there are empty bins.

[m-aguena]

Just to copy here what we discussed : we want to return NaN instead of 0 when there are empty bins.

done!

[combet]

Thank you @m-aguena - I just had a quick look, starting by running the demo_mock_ensemble notebook to check things out.

• I'm not sure I understand the output of cell 15. The legend says there should be dashed lines for "all table" but we only see solid lines in the figure.

• Something else changed and the covariance matrices later do not look like at all what they used to. Cf figures below when running from main and from this branch.

From main:

From this branch:

[combet]

Ah, I see, it's because maximum=nan in that cell (probably because of the new nan values in empty bins?), so the colorscale is not doing what it should...