Let's go for 2) for now

Codecov Report

❗ No coverage uploaded for pull request base (master@9fd3397). Click here to learn what that means.
The diff coverage is 81.64%.

@@            Coverage Diff            @@
##             master    #7218   +/-   ##
=========================================
  Coverage          ?   89.77%           
=========================================
  Files             ?      445           
  Lines             ?    80065           
  Branches          ?    12803           
=========================================
  Hits              ?    71880           
  Misses            ?     5375           
  Partials          ?     2810

What do you think @agramfort , @larsoner ?

I suspect codecov had an issue here.

There is a one-sided vs two-sided colormap problem here. Running:

import os
import mne
from mne.datasets import sample
data_path = sample.data_path()
sample_dir = os.path.join(data_path, 'MEG', 'sample')
subjects_dir = os.path.join(data_path, 'subjects')
fname_stc = os.path.join(sample_dir, 'sample_audvis-meg')
stc = mne.read_source_estimate(fname_stc, subject='sample')
initial_time = 0.1
with mne.viz.use_3d_backend('pyvista'):
    brain = stc.plot(subjects_dir=subjects_dir, initial_time=initial_time,
                     clim=dict(kind='value', pos_lims=[3, 6, 9]),
                     time_viewer=True)

I get something correct:

Until I hit "t", then I get something that scales in one-sided mode, which is incorrect:

Can you fix this, and check that this mode works both in two-sided/divergent mode (pos_lims) and one-sided (lims) mode?

The user given colormap values are now sent to _limits_to_control_points() which can detect both two-sided/divergent mode and one-sided mode.

There is also the shortcut u to restore the original clim.

I hope this fixed the issues you encountered @larsoner

If possible, I would prefer to merge this after #7219 and resolve the conflicts in here.

Looks like there are no merge conflicts, okay to test and merge @GuillaumeFavelier ?

The user given colormap values are now sent to _limits_to_control_points() which can detect both two-sided/divergent mode and one-sided mode.

I'm not sure if it should change the mode, though. If you initially plot in two-sided mode, it probably shouldn't switch to one-sided (and vice-versa). Currently it seems to, at least for the example where I plot data that are all positive in two-sided mode.

Passing the correct colormap in theory should prevent this, since we detect diverging colormaps there.

This also seems to work well in testing and I see no merge conflicts so I'll put it in. Thanks @GuillaumeFavelier