[BUG/Feature request] plotting dipoles gives unsatisfying and buggy results

Description of the problem

There doesn't seem to be a satisfying way to plot the dipoles obtained from dipole fit in MNE-Python. There are 4 options:

1. Orthoview: this is a bit low-contrast and it's not possible to traverse the slices. It's very difficult to identify anatomical landmarks as a result
   

2. Outlines: Similar problem as 1. If you get the dipoles at the wrong location, you can't interactively explore how far from the expected location you are.

3. Arrow/Sphere: seems buggy for me (see below)

Here is an example figure from a paper:

Why not have a vanilla option that resembles this with an option to interactively traverse slices? I saw this issue: https://github.com/mne-tools/mne-python/issues/2995, not sure if it's related. Please feel free to migrate the discussion there if relevant.

Steps to reproduce

import os.path as op
import numpy as np
import matplotlib.pyplot as plt

import mne
from mne.forward import make_forward_dipole
from mne.evoked import combine_evoked
from mne.simulation import simulate_evoked

from nilearn.plotting import plot_anat
from nilearn.datasets import load_mni152_template

data_path = mne.datasets.sample.data_path()
subjects_dir = data_path / 'subjects'
fname_ave = data_path / 'MEG' / 'sample' / 'sample_audvis-ave.fif'
fname_cov = data_path / 'MEG' / 'sample' / 'sample_audvis-cov.fif'
fname_bem = subjects_dir / 'sample' / 'bem' / 'sample-5120-bem-sol.fif'
fname_trans = data_path / 'MEG' / 'sample' / 'sample_audvis_raw-trans.fif'
fname_surf_lh = subjects_dir / 'sample' / 'surf' / 'lh.white'

evoked = mne.read_evokeds(fname_ave, condition='Right Auditory',
                          baseline=(None, 0))
evoked.pick_types(meg=True, eeg=False)
evoked_full = evoked.copy()
evoked.crop(0.07, 0.08)

# Fit a dipole
dip = mne.fit_dipole(evoked, fname_cov, fname_bem, fname_trans)[0]

# Plot the result in 3D brain with the MRI image.
dip.plot_locations(fname_trans, 'sample', subjects_dir, mode='arrow')

Link to data

No response

Expected results

Documentation says:

"Plot in 3D mode using PyVista with the given glyph type."

I am not sure what to expect since there is no example online ...

Actual results

I see this, not sure if it's an installation problem:

Additional information

In [11]: mne.sys_info()
Platform: macOS-11.5.2-x86_64-i386-64bit
Python: 3.9.7 | packaged by conda-forge | (default, Sep 2 2021, 17:58:46) [Clang 11.1.0 ]
Executable: /Users/mainak/anaconda3/envs/mne/bin/python
CPU: i386: 4 cores
Memory: 8.0 GB

mne: 1.3.dev0
numpy: 1.21.2 {unknown linalg bindings}
scipy: 1.7.1
matplotlib: 3.4.3 {backend=MacOSX}

sklearn: 0.24.2
numba: Not found
nibabel: 3.2.1
nilearn: 0.8.0
dipy: 1.4.1
openmeeg: Not found
cupy: Not found
pandas: 1.3.3
pyvista: 0.32.1 {OpenGL 4.1 INTEL-16.5.2 via Intel(R) HD Graphics 6000}
pyvistaqt: 0.5.0
ipyvtklink: 0.2.2
vtk: 9.0.1
qtpy: 1.11.0 {PyQt5=5.12.9}
ipympl: 0.9.1
pyqtgraph: 0.12.4
pooch: v1.6.0

mne_bids: Not found
mne_nirs: Not found
mne_features: Not found
mne_qt_browser: 0.2.5
mne_connectivity: Not found
mne_icalabel: Not found

Start by running the provided dipole.plot_locations reproduction with the sample dataset and reviewing its plotting documentation and implementation. Compare the current arrow, sphere, orthoview, and outlines modes with the reported screenshots; the issue does not define a single expected interaction or acceptance criteria, so the desired behavior needs clarification before implementation.