Supersample density maps

[eustermann@…]

To show x-ray and cryoEM meshes that appear smoother it is nice to use a smaller grid spacing, for instance, interpolating the map smoothly on a grid of half the grid spacing.

This makes large maps very large (e.g. 512^{3 becomes 1024}3) and can be slow. So the ability to do it just for subregions might be desirable.

[Tom Goddard]

This could be done with our current natural cubic spline code, interpolating each line of x-axis grid points to produce a finer sampling along x, then do the same along y and along z. (Not clear if there will be artifacts from interpolating the axes in order -- seems it would give slightly different result if a different order of axes was chosen.)

Maybe it is worth consider just supersampling the mesh without supersampling the density by dividing every triangle into 4 triangles and then smoothing.

[Tom Goddard]

Maybe a better way to do it than splining along each axis is to Fourier transform pad out with zeros, then Fourier transform back. I tried this in the past and believe it works well.