Check for feasibility in gen_candidates_scipy and error out for infea…

…sible candidates (#3335)

Summary:
Pull Request resolved: #3335

X-link: pytorch/botorch#2737

As titled. Previously, it was possible to return infeasible candidates to the user, with or without warnings alerting the user to the issue. This diff makes it so that the optimizer will error out when infeasible candidates are generated, so that the user can adjust the setup as needed.

Resolves pytorch/botorch#2708

Also includes a couple lint fixes in optimizer tests.

Reviewed By: esantorella

Differential Revision: D69314159

fbshipit-source-id: ccbc759274b2ae0fc2b38fa19d0e95abf35a2928

ax/modelbridge/tests/test_torch_moo_modelbridge.py

@@ -47,7 +47,7 @@
     get_non_monolithic_branin_moo_data,
     TEST_SOBOL_SEED,
 )
-from ax.utils.testing.mock import mock_botorch_optimize
+from ax.utils.testing.mock import mock_botorch_optimize, skip_fit_gpytorch_mll
 from ax.utils.testing.modeling_stubs import transform_1, transform_2
 from botorch.utils.multi_objective.pareto import is_non_dominated
 from pyre_extensions import assert_is_instance, none_throws
@@ -65,7 +65,7 @@ class MultiObjectiveTorchAdapterTest(TestCase):
         f"{STUBS_PATH}.BraninMetric.is_available_while_running",
         return_value=False,
     )
-    @mock_botorch_optimize
+    @skip_fit_gpytorch_mll
     def helper_test_pareto_frontier(
         self, _, outcome_constraints: list[OutcomeConstraint] | None
     ) -> None: