add single precision, add aa-vec-sl-soa kernel, updated doc

- Binaries have now a -dp or -sp suffix, depending on whether they have been
  compiled for double or single precision.

- New kernel for full array aa-vec-sl-soa added. Only one loop over the lattice
  used.

- Documentation has been updated, including how to build single precision
  binaries and performance graphs on various architectures.