fix allreduce tester (#385)

- use the reduce_psync buffers for synchronization in allreduce, not the
  barrier_psync.
- execute a wwg barrier after the allreduce operation. After internal
  discussion it was determined that it is required for correctness.

tests/functional_tests/team_reduction_tester.cpp

@@ -83,7 +83,7 @@ __global__ void TeamReductionTest(int loop, int skip, long long int *start_time,
   __shared__ rocshmem_ctx_t ctx;
   int wg_id = get_flat_grid_id();
 
-  rocshmem_wg_ctx_create(ctx_type, &ctx);
+  rocshmem_wg_team_create_ctx(team, ctx_type, &ctx);
 
   int n_pes = rocshmem_ctx_n_pes(ctx);