Add "explained variance" to dipole fitting GUI

[wmvanvliet]

When we fit dipoles, we currently display a GOF (goodness of fit) curve that is based on the covariance. However, what we really need is a "variance explained", as in: "how much of the sensor-level evoked timecourse is currently explained by the dipoles?" The current GOF measure is not the same, as the GOF will change if we load a different cov matrix while keeping the dipoles and evokeds the same. @larsoner

[wmvanvliet]

wait, that's exactly what the _compute_gof function does, except it also whitens the data, which is why the cov matters so much. Why are we whitening the data?