Custom filtering + cone-beam reconstruction

It would be good to combine any FBP filter (eg. Hilbert) and cone-beam reconstruction.

In theory it entails to also implement FDK pre-weighting on the nabu side. But in a first step we can assume that the cone is very small (i.e pre-weights=1).

This might fix the problematic memory usage of FDK (see #534 (closed) )

In principle FDK has the following steps:

• pre-weighting
• Parker weighting (for short scan)
• Filtering
• Backprojection with cone geometry