fix bug in gpu fft code, index calculation was wrong if number of maps was higher than number of maps per cufft call

Fix allocation of FFTW scratch space, move deallocation of memory allocated in configuration to cleanup function, get rid of MAX_ORIENT constant

merge compareRefMaps functions, create calc_logpro and update_prob function to remvoe code duplication

Add Option for asnychronous GPU processing, add test implementation with explicit SSE for CPU (currently disabled)