Merge pull request #4144 from CyrusBiotechnology/CYRUS-801-forcommons/nmer_speedup
Bugfix NmerRefEnergy. David Pearlman found that a large protein (it's a crispr) took thousands of seconds to run through our epitope scan protocol. Now it takes about 30 seconds, nearly all of which is initializing the score terms' const data from disk (and pose construction). Profiling revealed string construction was the problem. Turns out to be an unneeded deep copy operation. Thanks @decarboxy @roccomoretti @everyday847 for advice on how to profile. Tag @indigogo @BYachnin for awareness.