Tpetra: Improve thread scalability of CrsMatrix::fillComplete
Created by: mhoemmen
@trilinos/tpetra Epic: #796
This may or may not be worth elevating to its own epic, depending on pending performance results.
This should always refer to first fillComplete, since that's the only time when CrsMatrix::fillComplete does anything (other than possibly an all-reduce to check whether global assembly is necessary).