[Cluster] Improve message passing part.

This rewrites in part the message passing part, using in place gather, and collecting, rather than merging, the data of all threads.

neutral with a single thread per rank:
Score of new-2mpi-1t vs old-2mpi-1t: 789 - 787 - 2615  [0.500] 4191
Elo difference: 0.17 +/- 6.44

likely progress with multiple threads per rank:
Score of new-2mpi-36t vs old-2mpi-36t: 76 - 53 - 471  [0.519] 600
Elo difference: 13.32 +/- 12.85

src/thread.h

@@ -78,9 +78,8 @@ public:
 #ifdef USE_MPI
   struct {
       Mutex mutex;
-      Cluster::TTSendBuffer<Cluster::TTSendBufferSize> buffer = {};
-      size_t counter = 0;
-  } ttBuffer;
+      Cluster::TTCache<Cluster::TTCacheSize> buffer = {};
+  } ttCache;
 #endif
 };