ACRC Improvements

[harrywaugh]

Summary of Changes:

• Added some restrict and const qualifiers so that the compiler is better at auto-vectorizing the propagate loop.
• Edited the Makefile so that pb_mpi can be build with Intel compiler, this can be done using make INTEL=1. (Faster than GNU currently). To compile on Intel, some missing return statements were also inserted to preserve the correctness of the output trace file.
• Removed explicit MPI_Barriers, which then idenitifes MPI_Bcast as a major bottleneck with more cores.
• Added benchmark report.