Performance tuning for inter-node communication (#66)

This PR addresses two issues:
 - reduce the number of contexts supported by the host-interface by
   default to 1, we are not using those at the moment, and hence
   we now create fewer MPI_Win at the startup
 - introduces a micro-sleep in RO progress engine in case there are no
   pending requests. This leads significant performance improvements
   observed for inter-node communication with THor2 NICs.

[ROCm/rocshmem commit: 12561783de]

projects/rocshmem/src/host/host.hpp

@@ -341,7 +341,7 @@ class HostInterface {
   /**
    * @brief Max number of contexts for the application
    */
-  int max_num_ctxs_{40};
+  int max_num_ctxs_{1};
 
   /**
    * @brief Pool of HostContexWindowInfos