Speed up dipole fit rendering

[larsoner]

@wmvanvliet this about doubles FPS on my machine by (with the help of Opus 5 drafting and iteratively testing for speed comparisons):

1. Using mpl blitting for the time plot
2. Using nearest smoothing steps (should be fine for these brain plots)
3. Speeding up overlay computation using contiguity
4. Only recomputing overlay parts when needed
5. Avoiding recomputation of the contour algorithm

Many of these will also speed up plotting in brain, which is nice!

Feel free to give it a whirl and if it works for you we should be good to go

[wmvanvliet]

Can we have the blitting code for other matplotlib figures as well? Now that we're adding more interactivity and ui-event integrations, we have more and more "time cursors" in figures. For example, plot_evoked_topo has a time cursor that is pretty slow.

[drammock]

Can we have the blitting code for other matplotlib figures as well? Now that we're adding more interactivity and ui-event integrations, we have more and more "time cursors" in figures. For example, plot_evoked_topo has a time cursor that is pretty slow.

blitting should be fine for any figure where the axes ranges don't change as a result of interaction. It won't help (much) for, e.g, the raw/epochs data browser.