test_optimize_kl: use non-static ncg

See: #426

test/test_re/test_optimize_kl.py

@@ -222,7 +222,9 @@ def test_optimize_kl_sample_consistency(
         point_estimates=point_estimates,
         draw_linear_kwargs=draw_linear_kwargs,
         nonlinearly_update_kwargs=dict(minimize_kwargs=minimize_kwargs),
-        kl_kwargs=dict(minimize_kwargs=dict(name="M", maxiter=0)),
+        kl_kwargs=dict(
+            minimize=jft.optimize._newton_cg, minimize_kwargs=dict(name="M", maxiter=0)
+        ),
         sample_mode=sample_mode,
         kl_jit=False,
         residual_jit=False,