Vector geometries

[kohr-h]

We should have geometries that simply wrap the ASTRA vec geometries, for full flexibility regarding positions of everything. If we want to cram this into the framework of curves and all that, we can simply say that source and detector move along piecewise linear paths.
What this will need in any case is a variant of the ray transform whose range doesn't perform any weighting in the motion parameters.

I guess this is related to #152, definitely to #285.

[adler-j]

Also see #456 and #457.

For the motion parameters I agree.

[kohr-h]

I added vector geometries as a PR into the #968 branch. Works well and makes #456 and #457 quite simple to address.