The collection of statistics for concurrent refinement is somewhat different between refinment threads and mutator threads, making some of the code a bit awkward. In particular, determining the number of dirty cards that were logged has some problems. Also, the refinement rate doesn't include mutator refinement.

Obtaining info about concurrent refinement statistics is also awkward, as it's included in the RemsetSummary, which requires additional command line options to enable, and generating a RemsetSummary is somewhat expensive when all we want is concurrent refinement statistics.

We should collect more consistent data for the refinement threads and mutator refinement, and provide easier access to that information.