Update RCCL/MSCCL work FIFO depth to 256K (#1091)

[ROCm/rccl commit: c5ab37211b]

projects/rccl/src/init.cc

@@ -443,7 +443,7 @@ NCCL_PARAM(AggChannelSize, "AGG_CHANNEL_SIZE", -2);
 NCCL_PARAM(DisableGraphHelper, "GRAPH_HELPER_DISABLE", 0);
 // GDRCOPY support: FIFO_ENABLE when enabled locates a workFifo in CUDA memory
 NCCL_PARAM(GdrCopyFifoEnable, "GDRCOPY_FIFO_ENABLE", 1);
-NCCL_PARAM(WorkFifoDepth, "WORK_FIFO_DEPTH", 64<<10);
+NCCL_PARAM(WorkFifoDepth, "WORK_FIFO_DEPTH", 256<<10);
 enum ncclLaunchMode ncclParamLaunchMode;