oshmem/ucx: improves spml ucx performance for multi-threaded

applications.

Improves multi-threaded performance by adding the option to create
multiple ucx workers in threaded applications.

Co-authored with:
Artem Y. Polyakov <artemp@mellanox.com>,
Manjunath Gorentla Venkata <manjunath@mellanox.com>

Signed-off-by: Tomislav Janjusic <tomislavj@mellanox.com>

Author: Tomislav Janjusic

oshmem/mca/atomic/ucx/atomic_ucx_cswap.c

@@ -50,6 +50,6 @@ int mca_atomic_ucx_cswap(shmem_ctx_t ctx,
         mca_spml_ucx_remote_op_posted(ucx_ctx, pe);
     }
 
-    return opal_common_ucx_wait_request(status_ptr, ucx_ctx->ucp_worker,
+    return opal_common_ucx_wait_request(status_ptr, ucx_ctx->ucp_worker[0],
                                         "ucp_atomic_fetch_nb");
 }

oshmem/mca/atomic/ucx/atomic_ucx_module.c

@@ -80,7 +80,7 @@ int mca_atomic_ucx_fop(shmem_ctx_t ctx,
                                      op, value, prev, size,
                                      rva, ucx_mkey->rkey,
                                      opal_common_ucx_empty_complete_cb);
-    return opal_common_ucx_wait_request(status_ptr, ucx_ctx->ucp_worker,
+    return opal_common_ucx_wait_request(status_ptr, ucx_ctx->ucp_worker[0],
                                         "ucp_atomic_fetch_nb");
 }