refactor: refactor the entire file organization structure.

[magicheng0816]

No description provided.

[gemini-code-assist]

Code Review

This pull request introduces a large-scale refactoring of the service's file organization structure. The changes are extensive, involving renaming of core components (Scheduler to Executor, http_service to api_service, rpc_service to instance_service, and load_balance_policy to scheduling_policy), moving files to a more modular directory structure, and updating CMakeLists.txt files and include paths accordingly. The refactoring appears to be consistent and well-executed. I've identified one critical issue related to graceful shutdown that was present in the old code and has been carried over to the new structure. Fixing it will improve the service's robustness.

[gemini-code-assist]

The current shutdown_handler calls exit(1), which terminates the process immediately and prevents graceful shutdown. This can lead to resource leaks or inconsistent state, as cleanup code like master.stop() is never executed. To enable graceful shutdown, the signal handler should set an atomic flag to signal the main loop to terminate. This allows master.stop() to be called, which in turn ensures that server threads are properly joined.

Suggested change

	void shutdown_handler(int signal) {
	LOG(WARNING) << "Received signal " << signal << ", stopping master...";
	exit(1);
	}
	void shutdown_handler(int signal) {
	LOG(WARNING) << "Received signal " << signal << ", stopping master...";
	g_signal_received.store(1, std::memory_order_relaxed);
	}