Squueze out more performance from Regular Lattice

templatifying the dimensionality we can et rid of std::vector allocations and deallocations.