Add Nakama metrics and structured AI decision observability

[JOY]

Project: Second Spawn
Area: Nakama
Milestone: Foundation
Priority: P1
Size: M
Review Gate: Local Review
Verification: Backend Tests

Goal

Make Nakama and AI-agent behavior observable before larger autonomous NPC tests.

Scope

• Document and enable Nakama metrics export for local/staging operations.
• Add structured logs for agent decisions: actor id, source, source reason, allowed actions, selected action, target id, validation result, model, and latency.
• Track rate-limit, token-budget, storage-conflict, reward-claim, and BodyTime mutation events.
• Preserve player privacy and avoid logging provider secrets.

Acceptance Criteria

• Local/staging operators can see model vs fallback rate.
• Important AI and economy validation branches produce structured logs.
• No API keys or private prompt payloads are logged.