we should use fftw_malloc to allocate memory used for ffts (since we do not consider correct aliasing with new), anyhow, the conv_map variable is only used for FFTAlgo=0 and should not be allocated otherwise

fix sum and sumsquare calculation: for the non FFT algorithm sum was calculated twice before (stupid mistake), for the FFT algorithm there is actually a dependency for the sumsquare calculation on whether NumberPixels is odd or even. Although, the effect is very small for normal maps since high order fourier coefficients are small then. Anyhow, should be correct now

improve debug output, use openmp during initialization where applicable, move fft scrach spaces to param class

make FFTALGO default, add MPI initialization, move allocation of sumsquare pointer to precalculate function