Merge pull request #4766 from RosettaCommons/roccomoretti/vtune_1
Some simple efficiency tweaks.
I was playing around with the Intel VTune Profiler (on relax runs), and noticed a few "high spots" which could be addressed without too much restructuring.
Mainly these were in reducing the number of memory allocations, though there were a few runtime-related issues. I don't think there's substantial performance improvements here, but I think all of the changes fall into the category of "not worse", either in runtime or comprehensibility.