Raise application-level exceptions for some failure scenarios. (#2429)

* Raise application level exception for actor methods that can't be executed and failed tasks.

* Retry task forwarding for actor tasks.

* Small cleanups

* Move constant to ray_config.

* Create ForwardTaskOrResubmit method.

* Minor

* Clean up queued tasks for dead actors.

* Some cleanups.

* Linting

* Notify task_dependency_manager_ about failed tasks.

* Manage timer lifetime better.

* Use smart pointers to deallocate the timer.

* Fix

* add comment

src/ray/object_manager/object_manager.h

@@ -38,7 +38,7 @@ struct ObjectManagerConfig {
   int max_receives;
   /// Object chunk size, in bytes
   uint64_t object_chunk_size;
-  /// The stored socked name.
+  /// The store socket name.
   std::string store_socket_name;
   /// The time in milliseconds to wait until a Push request
   /// fails due to unsatisfied local object. Special value:
@@ -101,8 +101,8 @@ class ObjectManager : public ObjectManagerInterface {
   ///
   /// \param object_id The object's object id.
   /// \param client_id The remote node's client id.
-  /// \return Status of whether the push request successfully initiated.
-  ray::Status Push(const ObjectID &object_id, const ClientID &client_id);
+  /// \return Void.
+  void Push(const ObjectID &object_id, const ClientID &client_id);
 
   /// Pull an object from ClientID. Returns UniqueID asociated with
   /// an invocation of this method.