Sync the variable with previous if

include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp

@@ -1076,7 +1076,7 @@ struct GridwiseGemmMultipleDWelfordFirstHalf_xdl_cshuffle
                     // Stride of count is [0, 1]. Only the first row in count[0, 0:nblock] need
                     // to be written.
                     if(i == 0 && block_work_idx[I0] == 0 &&
-                       post_shuffle_thread_data_idx_begin[I0] == 0)
+                       post_shuffle_thread_cluster_idx[I0] == 0)
                     {
                         auto count_thread_copy_vgpr_to_global = ThreadwiseTensorSliceTransfer_v1r3<
                             int32_t,