Mesh optimizations

Created by: dominikkiese

The scanning routine for dynamical adjustment of the frequency meshes is optimized and yields quite good results down to cutoffs Lambda / |J| ~ 0.05, which has therefore been made the default value for final in the launch! function. Furthermore, unnecessary index searches in higher loops and the BSEs were removed. There have been some additional changes such as proper normalization of the cutoff (w.r.t. the spin coupling), vectorization of the integration routine and implementation of Richardson extrapolation for quadrature results in subdomains. Can you look over this and check that especially the scanning, normalization and extrapolation have not messed things up @LoctusTM / @lgresista ?