Merge pull request #1268 from RosettaCommons/roccomoretti/icc_pseudofix
Change RotamericSingleResidueDunbrackLibrary to avoid unnecessary intermediate data structure.
This should fix the current issue with fa_dun score differences between GCC and ICC.
This doesn't, however, necessarily address the underlying issue of why ICC was breaking the code when optimizing. I've (separately) put together a test case to pass along to Intel as a possible bug report.