Optimize: shift `e1%out_cont%polyline` instead of shifting
`e2%speed_cont%polyline` and `e2%out%polyline`. This can save one call
to `copy_shift`.