Fix NVHPC warnings in native GPU build sources

Remove unused sync_voltage_and_rhs_to_device helper, mark
solve_interleaved2_loop_body as routine gang so inner loop vector
directives are valid under the outer gang parallel region, and use
the nodiscard ensure_on_device return value in the device_state test.

Author: nrnhines

src/coreneuron/permute/cellorder.cpp

@@ -619,7 +619,7 @@ static void bksub_interleaved(NrnThread* nt,
     }
 }
 
-nrn_pragma_acc(routine vector)
+nrn_pragma_acc(routine gang)
 static void solve_interleaved2_loop_body(NrnThread* nt,
                                          double* vec_a,
                                          double* vec_b,

src/neuron/gpu/sync.cpp

@@ -84,22 +84,6 @@ void sync_matrix_arrays_to_host(NrnThread& nt) {
 #endif
 }
 
-void sync_voltage_and_rhs_to_device(NrnThread& nt) {
-#if defined(NRN_ENABLE_GPU)
-    if (!nt.compute_gpu || nt.end <= 0) {
-        return;
-    }
-    auto* const vec_v = nt.node_voltage_storage();
-    auto* const vec_rhs = nt.node_rhs_storage();
-    nrn_pragma_acc(update device(vec_v [0:nt.end], vec_rhs [0:nt.end]) if (nt.compute_gpu)
-                       async(nt.stream_id))
-    nrn_pragma_omp(target update to(vec_v [0:nt.end], vec_rhs [0:nt.end]) if (nt.compute_gpu))
-    nrn_pragma_acc(wait(nt.stream_id))
-#else
-    (void) nt;
-#endif
-}
-
 }  // namespace
 
 void sync_before_vecplay(NrnThread& nt) {

test/unit_tests/gpu/device_state.cpp

@@ -122,7 +122,8 @@ TEST_CASE("invalidate_device_state clears GPU mirrors", "[gpu][device_state]") {
     neuron::container::Node::storage nodes{};
     auto node_token = nodes.issue_frozen_token();
     neuron::model_sorted_token sorted{cache, std::move(node_token)};
-    ensure_on_device(sorted);
+    device_token const& token = ensure_on_device(sorted);
+    REQUIRE(token.is_on_device());
     REQUIRE(detail::is_on_device_for_testing());
     invalidate_device_state();
     REQUIRE_FALSE(detail::is_on_device_for_testing());