prov/efa: Fix bugs in domain queue lists iteration

When hitting EAGAIN, we should continue instead of
break, because now the list can contain opes for
other eps, which may not have resource exhaustion.

When hitting error, we should still move forward
after writing the txe/rxe error.

Signed-off-by: Shi Jin <sjina@amazon.com>

prov/efa/src/efa_domain.c

@@ -468,14 +468,14 @@ void efa_domain_progress_rdm_peers_and_queues(struct efa_domain *domain)
 
 		ret = efa_rdm_ep_post_handshake(peer->ep, peer);
 		if (ret == -FI_EAGAIN)
-			break;
+			continue;
 
 		if (OFI_UNLIKELY(ret)) {
 			EFA_WARN(FI_LOG_EP_CTRL,
 				"Failed to post HANDSHAKE to peer %ld: %s\n",
 				peer->efa_fiaddr, fi_strerror(-ret));
 			efa_base_ep_write_eq_error(&peer->ep->base_ep, -ret, FI_EFA_ERR_PEER_HANDSHAKE);
-			return;
+			continue;
 		}
 
 		dlist_remove(&peer->handshake_queued_entry);
@@ -600,7 +600,7 @@ void efa_domain_progress_rdm_peers_and_queues(struct efa_domain *domain)
 					break;
 
 				efa_rdm_txe_handle_error(ope, -ret, FI_EFA_ERR_PKT_POST);
-				return;
+				continue;
 			}
 		}
 	}