Indexing random sphere

This includes all my bugfixes of the last merge request, which are necessary to load multiple nexus files.

Furthermore, it makes more explicit the flaw in the current version: The unseeded random number generation in the previous version of FFTIndexing.cpp always gave back the same number, namely 0.131538 (for our version of libstdc++), and only with this number does TestAutoIndexer.py work.

Instead of the non-working implementation of randomness, this number is now hard-coded to make the problem obvious, which is still better than "randomly" risking a dependency on what libstdc++ gives us back as the first unseeded value. The actual random-number generation of my previous merge request can be re-enabled with RANDOMISE_FFTINDEXING_SPHERE.

This random number only shifts the generated points on the sphere used for fft indexing, the sphere itself is okay in any of the cases (see new test TestFFTIntexingSphere.cpp). So TestAutoIndexer.py should work with any random shift of the sphere vertices with n_vertices large enough. That it does not is an indicator of another possibly more serious problem.