fix(envs): use forkserver context and close envs in test to prevent deadlock

AsyncVectorEnv with default fork context leaks worker processes between
test_policy parametrized cases; subsequent env creation deadlocks because
new forked workers inherit stale pipe FDs from previous test's leaked workers.

- configs.py: pass context="forkserver" to AsyncVectorEnv (matches _LazyAsyncVectorEnv)
- test_policies.py: call close_envs(envs) at end of test_policy to clean up workers

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/lerobot/envs/configs.py

@@ -109,12 +109,17 @@ class EnvConfig(draccus.ChoiceRegistry, abc.ABC):
         def _make_one():
             return gym.make(self.gym_id, disable_env_checker=self.disable_env_checker, **self.gym_kwargs)
 
+        extra_kwargs: dict = {}
+        if env_cls is gym.vector.AsyncVectorEnv:
+            extra_kwargs["context"] = "forkserver"
         try:
             from gymnasium.vector import AutoresetMode
 
-            vec = env_cls([_make_one for _ in range(n_envs)], autoreset_mode=AutoresetMode.SAME_STEP)
+            vec = env_cls(
+                [_make_one for _ in range(n_envs)], autoreset_mode=AutoresetMode.SAME_STEP, **extra_kwargs
+            )
         except ImportError:
-            vec = env_cls([_make_one for _ in range(n_envs)])
+            vec = env_cls([_make_one for _ in range(n_envs)], **extra_kwargs)
         return {self.type: {0: vec}}
 
     def get_env_processors(self):

tests/policies/test_policies.py

@@ -31,7 +31,7 @@ from lerobot.datasets.factory import make_dataset
 from lerobot.datasets.feature_utils import dataset_to_policy_features
 from lerobot.datasets.utils import cycle
 from lerobot.envs.factory import make_env, make_env_config
-from lerobot.envs.utils import preprocess_observation
+from lerobot.envs.utils import close_envs, preprocess_observation
 from lerobot.optim.factory import make_optimizer_and_scheduler
 from lerobot.policies.act.configuration_act import ACTConfig
 from lerobot.policies.act.modeling_act import ACTTemporalEnsembler
@@ -224,6 +224,8 @@ def test_policy(ds_repo_id, env_name, env_kwargs, policy_name, policy_kwargs):
     # Test step through policy
     env.step(action)
 
+    close_envs(envs)
+
 
 # TODO(rcadene, aliberts): This test is quite end-to-end. Move this test in test_optimizer?
 def test_act_backbone_lr():