ENH topomap validation and bad channels

When using mne to plot topomaps computed with some external methods, it highlights a channel that appears to be "bad", in the sense that the topomap is discontinuous:

Here is a snippet reproducing this topomap with files from numpy_files.zip:

import numpy as np
import mne

pos = np.load("pos.npy")
u_hat = np.load("spatial_map.npy")

mne.viz.plot_topomap(u_hat, pos)

I would be interested in having a utility function returning the spatial regularity of a topomap and the potential bad channels. And I think it would be better in mne as you need to correctly parse the position of each sensor to compute the spatial differential.

Start with the issue's reproducer using pos.npy, spatial_map.npy, and mne.viz.plot_topomap. Read the topomap visualization entry point and determine how sensor positions are parsed. Done should include a decided utility API that measures spatial regularity and identifies potential bad channels, with coverage for the supplied example.