Do a more detailed performance assessment to identify hotspots

(Task created in NCAR Teamwork 2015-11-11, but still relevant.)

We should do a more detailed performance assessment to identify hotspots. This should include more detailed timers.

In addition, if possible, we should use a tool that identifies cache efficiency, to determine where we are incurring large costs due to cache inefficiencies. (Bill: I feel this is an issue due to (1) the large cost we see of CNZero, and (2) the large cost added when adding 0-weight, inactive columns.)