[2.0] Add Generals.io bot arena task

[joyemang33]

Summary

Add a new Frontier-CS 2.0 generals_io_bot task: agents submit patches to a Generals.io bot implementation and are evaluated in a Harborized arena against multiple baseline bots.

This PR includes:

• A self-contained 2.0/problems/generals_io_bot task with task config, statement, evaluator, reference patch, Harbor app files, and upstream license credit.
• Task-specific agent/judge Docker images for the Generals runtime and bundled bot dependencies.
• Patch-based submission flow where an empty patch is a valid baseline submission.
• Full-baseline intermediate and final evaluation, scored by win rate plus a speed tiebreaker.
• Minimal 2.0 judge/template support needed by this task: empty submissions, final reruns in a child process, and task-configurable async start method.

Please read CONTRIBUTING.md before submitting.

Type of Change

• New research problem
• New algorithmic problem
• New Frontier-CS 2.0 problem
• Bug fix
• Documentation update
• Other:

Testing

• Built the Generals.io task-specific agent and judge Docker images.
• Generated the Harbor task through the Frontier-CS 2.0 adapter.
• Ran direct evaluator smoke tests for empty/reference-style submissions.
• Ran judge HTTP submission smoke tests.
• Ran a full Harbor trial with Codex:

{
  "reward": 0.3664351851851852,
  "score": 36.64351851851852,
  "score_unbounded": 36.64351851851852,
  "trial_status": "scored",
  "agent_status": "AgentTimeoutError",
  "successful_submissions": 2
}