Rename rollout_engines and rank to reflect semantics

[fzyzcjy]

No description provided.

[gemini-code-assist]

Code Review

This pull request renames the rollout_engines variable to new_engines in server_group.py and adds a maintenance note in addr_allocator.py. The review feedback correctly identifies that renaming the rank variable to index in the engine initialization loop reduces semantic clarity, as the value represents a global rank in a distributed computing context.

[gemini-code-assist]

Renaming rank to index here is less descriptive and potentially misleading. In distributed computing contexts, "rank" is the standard term for a process identifier within a group, whereas "index" typically refers to a position in a sequence (like a 0-based list index). Since the value being unpacked from new_engines is the global_rank (as seen in line 120), it is better to maintain the name rank or global_rank to reflect its actual semantics and maintain consistency with the rest of the codebase.

Suggested change

	**addr_and_ports[index],
	router_ip=self.router_ip,
	router_port=self.router_port,
	)
	for rank, engine in rollout_engines
	for index, engine in new_engines
	**addr_and_ports[rank],
	router_ip=self.router_ip,
	router_port=self.router_port,
	)
	for rank, engine in new_engines