Take units (m or mm) into account when showing fieldmaps on top of brains

[wmvanvliet]

Plotting evoked fieldlines on top of distributed source estimates is very cool:

But, mne.viz.Brain can use either units="m" or units="mm". The plot_evoked_field function always assumed it would be kept at the default units="mm". This PR makes it actually pay attention to the unit. Another bug fixed is that when plotting in millimeters, the helmet surface would be modified in place: surf["rr"] *= 1000. This causes trouble as now the surface will forever be upscaled and should you plot it again, it will be upscaled again. In this PR a deepcopy() is made first.

[drammock]

windows pip-pre job: odd failure in test_search_light:

mne/decoding/tests/test_search_light.py::test_search_light_basic ERROR   [ 21%]
C:\hostedtoolcache\windows\Python\3.12.8\x64\Lib\site-packages\pyarrow\compute.py:211: RuntimeWarning: Python binding for RankQuantileOptions not exposed
  warnings.warn("Python binding for {} not exposed"
C:\hostedtoolcache\windows\Python\3.12.8\x64\Lib\site-packages\pyarrow\compute.py:211: RuntimeWarning: Python binding for RankQuantileOptions not exposed
  warnings.warn("Python binding for {} not exposed"

ubuntu conda 3.12 had a segfault during test_3d.py, not sure if related.

[larsoner]

windows pip-pre job: odd failure in test_search_light:

Ugh probably some sklearn thing, I'll push a quick ignore

[larsoner]

Marking for merge-when-green, thanks in advance @wmvanvliet

(I think the conda segfault is unrelated, I think we just get those sometimes but we'll see...)

[wmvanvliet]

Thanks for the great debugging @larsoner!