cpp mt benchmark should be modified

[FAKERINHEART]

Your cpp mt benchmark does not take advantage of the grpc async modes.
There is always only one CallData in one thread, which will have adverse side effects in grpc-server receiving packets.
You can pre-create more than one CallData(tag)s registered into the server.

[LesnyRumcajs]

I believe the code was taken directly from the C++ gRPC examples. Did those change? If not, should they change if you have a more optimal solution?