Add gather, scatter and alltoall collectives

Introducing 3 new APIs:
ncclResult_t  ncclGather(const void* sendbuff, void* recvbuff, size_t sendcount,
    ncclDataType_t datatype, int root, ncclComm_t comm, hipStream_t stream);
ncclResult_t  ncclScatter(const void* sendbuff, void* recvbuff,
    size_t recvcount, ncclDataType_t datatype, int root, ncclComm_t comm,
    hipStream_t stream);
ncclResult_t  ncclAllToAll(const void* sendbuff, void* recvbuff, size_t count,
    ncclDataType_t datatype, ncclComm_t comm, hipStream_t stream);

Only out of place operation is supported.
Preprocessor symbol RCCL_GATHER_SCATTER=1 indicates API availibility.
By default the APIs launche RCCL kernel implementation, which can be disabled by
RCCL_ALLTOALL_KERNEL_DISABLE=1. Then the APIs use wrapper around ncclSend and ncclRecv.


[ROCm/rccl commit: e80e29573c]

projects/rccl/src/include/devcomm.h

@@ -23,8 +23,8 @@
 #endif
 
 #define NCCL_NUM_FUNCTIONS 5 // SendRecv not included for now
-typedef enum { ncclCollBroadcast, ncclCollReduce, ncclCollAllGather, ncclCollReduceScatter, ncclCollAllReduce, ncclCollSendRecv} ncclFunc_t;
-extern const char* ncclFuncStr[NCCL_NUM_FUNCTIONS];
+typedef enum { ncclCollBroadcast, ncclCollReduce, ncclCollAllGather, ncclCollReduceScatter, ncclCollAllReduce, ncclCollGather, ncclCollScatter, ncclCollAllToAll, ncclCollSendRecv} ncclFunc_t;
+extern const char* ncclFuncStr[NCCL_NUM_FUNCTIONS+3];
 
 #define NCCL_NUM_ALGORITHMS 3 // Tree/Ring/CollNet
 #define NCCL_ALGO_TREE 0