MNT: Add brain_gc fixture to test_process_clim_plot

[GuillaumeFavelier]

This PR adds the brain_gc fixture in test_process_clim_plot (I missed it in #8427) and fixes the issues with memory:

_______________________________________________________________________________ ERROR at teardown of test_process_clim_plot[pyvista] _______________________________________________________________________________
mne/conftest.py:522: in brain_gc
    assert len(bad) == 0, 'VTK objects linger:\n' + '\n'.join(bad)
E   AssertionError: VTK objects linger:
E     vtkXOpenGLRenderWindow
E     vtkGenericRenderWindowInteractor
E   assert 2 == 0
E     +2
E     -0

It's also important for #8335 (that's how I discovered the issue in the first place)

[larsoner]

I don't fully understand -- doesn't this indicate that the close_all that happens at the end of this test is somehow not working? How do we get a figure set/created in a way that our close-all fixture will not close it?

[GuillaumeFavelier]

We keep an internal reference to mimic mayavi's behaviour when figure (int) is used.

[larsoner]

See follow-up comment -- I think this means close_all needs to be updated, or that we need to update our fixture to call another close_all for the case of int-passing or something.

[larsoner]

To put it differently, whenever you add brain_gc to a test, any failures should be fixed in _brain.py, not in the test itself -- this is because brain_gc does a close_all sort of operation, and this should close all figures. If it doesn't, it suggests some problem with our close_all that should be fixed.

[larsoner]

LGTM feel free to merge once CIs are green @GuillaumeFavelier