Fix bandwidth selection in sim_proc_many_pops

[philliplab]

This is more complex because there is interplay between

geom_smooth and density

that is tricky.

Also, most of the estimation should be treated as if for a discrete distributoins? Way around this is to just increase the sequence length?

[philliplab]

I need a small investigation into this.

What cutoff can be designed to decide if the density estimation should be discrete or continuous?

[philliplab]

See also design vignette