Add bfloat16 support in RCCL

Preprocessor symbol RCCL_BFLOAT16 is used as feature indicator

src/include/core.h

@@ -48,6 +48,7 @@ static __inline__ int ncclTypeSize(ncclDataType_t type) {
     case ncclUint8:
       return 1;
     case ncclFloat16:
+    case ncclBfloat16:
       return 2;
     case ncclInt32:
     case ncclUint32: