Add AllGather LL128 multi-node tuning and include LL cutoff points in tuning models (#1618)

* Enable LL/LL128 cutoff points in tuning models

* Initializing ll/ll128 model cutoffs for MI300

* Use RCCL_LL_LIMITS_UNDEFINED

---------

Co-authored-by: PedramAlizadeh <pmohamma@amd.com>

[ROCm/rccl commit: 4be06f04d8]

projects/rccl/src/include/comm.h

@@ -517,6 +517,7 @@ struct ncclComm {
   float bandwidths[NCCL_NUM_FUNCTIONS][NCCL_NUM_ALGORITHMS][NCCL_NUM_PROTOCOLS];
   float ringbdw[NCCL_NUM_FUNCTIONS][NCCL_NUM_PROTOCOLS];
   int maxThreads[NCCL_NUM_ALGORITHMS][NCCL_NUM_PROTOCOLS];
+  uint64_t minMaxLLRange[NCCL_NUM_FUNCTIONS][NCCL_NUM_PROTOCOLS - 1][2];
 
   /* This attribute can indicate the states of communicators and return code of
   * asynchronous NCCL operations. */
@@ -599,7 +600,7 @@ struct ncclComm {
   struct ncclKernelPlanner planner;
 
   hipStream_t sideStream; // [RCCL] Cached non-captured stream
-  
+
   cudaMemPool_t memPool;
   // Queue of events and associated callbacks for cleaning up asynchronous work.
   // Using this is preferable to using CUDA host callbacks because host callbacks