refactor(pi052): trim PR — remove say tool, debug gates, dead code; move runtime

Cleanup pass over the language-support PR to cut LOC and scope creep.

Removals:
- SayTool + tools/ package (registry, Tool protocol, [tools] extra) and the
  runtime's tool-dispatch path. Kept <say> training supervision and inference
  stripping so speech-annotated datasets still train.
- WeightedEpisodeAwareSampler + VQA oversampling wiring
  (_build_vqa_oversample_weights, vqa_target_fraction) — training uses plain
  EpisodeAwareSampler again.
- Debug env-gates PI052_DEBUG_TENSORS, PI052_SUBTASK_USE_TASK, EVAL_TASK_OVERRIDE.
- Dead code: broken _tp._DUMP_BUDGET block, unused imports (copy/Tensor,
  RevisionNotFoundError, LeRobotDataset, os), messages_for_vqa, steps.py shim
  (modeling imports pi052_adapter directly), duplicated _emit, builtins.type[T].

Moves:
- Policy-agnostic runtime -> src/lerobot/runtime/ (LanguageConditionedRuntime +
  adapter Protocol + state); pi052 keeps only its adapter + CLI. Tests -> tests/runtime/.

Other:
- Compacted verbose AI-authored comments/docstrings across pi052 (kept the
  hard-won DDP / barrier-timeout / reduce-max / VQA-routing notes).
- Relocated LM-head prediction debug helper to pi052/debug_utils.py.
- Fixed test_render_messages: assert task-fallback render (current behavior)
  instead of the stale no-op expectation.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

tests/policies/pi052/test_pi052_runtime_adapter.py

@@ -1,7 +1,7 @@
 from types import SimpleNamespace
 
-from lerobot.policies.language_conditioned import RuntimeState
 from lerobot.policies.pi052.inference.pi052_adapter import PI052PolicyAdapter, split_plan_and_say
+from lerobot.runtime import RuntimeState
 
 
 def test_pi052_adapter_builds_recipe_prompts_from_runtime_state():
@@ -28,13 +28,11 @@ def test_pi052_adapter_builds_recipe_prompts_from_runtime_state():
     ]
 
 
-def test_pi052_adapter_parses_say_tool_calls_and_plan_text():
+def test_pi052_adapter_strips_say_markers_from_plan_text():
     adapter = PI052PolicyAdapter(policy=object())
     text = "Move to the sink. <say>heading to the sink</say>"
 
     assert split_plan_and_say(text) == ("Move to the sink.", "heading to the sink")
-    assert adapter.parse_tool_calls(text)[0].name == "say"
-    assert adapter.parse_tool_calls(text)[0].arguments == {"text": "heading to the sink"}
     assert adapter.plan_from_text(text) == "Move to the sink."
 
 
@@ -48,7 +46,6 @@ def test_pi052_runtime_cli_smoke_does_not_load_model(monkeypatch):
         "_load_policy_and_preprocessor",
         lambda policy_path, dataset_repo_id: (fake_policy, None, None, None),
     )
-    monkeypatch.setattr(runtime_cli, "_build_tools", lambda no_tts, tts_voice: {})
     monkeypatch.setattr(runtime_cli, "_run_repl", lambda runtime, initial_task, max_ticks: 0)
 
     assert runtime_cli.main(["--policy.path=fake", "--no_robot", "--task=clean", "--max_ticks=0"]) == 0