SIENTIAPDE-1314

Enhance thread safety in MLFlowRepository model caching

- Introduced a reentrant lock to synchronize access to the model cache, ensuring thread safety during cache checks and updates.
- Updated the cache management logic to acquire the lock when checking for existing models and when updating the cache after downloading a new model.
- Reduced the maximum cached workflows in the worker configuration for improved resource management.

laborious/worker/worker.py

@@ -152,7 +152,7 @@ async def main():
             max_concurrent_workflow_tasks=50,
             max_concurrent_activities=50,
             max_concurrent_local_activities=50,
-            max_cached_workflows=200,
+            max_cached_workflows=2,
             workflow_task_poller_behavior=PollerBehaviorAutoscaling(),
             activity_task_poller_behavior=PollerBehaviorAutoscaling(),
         ),
@@ -164,7 +164,6 @@ async def main():
                 # MLFlow
                 activities.request_predict,
                 activities.request_transform,
-                activities.query_to_minio,
                 # Gates
                 activities.input_gate,
                 activities.mlflow_response_gate,