AI Gateway optimization

[chitcommit]

Description\nOptimize AI Gateway for latency and cost efficiency.\n\nTasks\n- [ ] Implement response caching\n- [ ] Add request batching\n- [ ] Optimize prompt templates\n- [ ] Add rate limiting per tenant\n- [ ] Monitor token usage\n\nGoals\n- < 500ms p95 latency\n- 30% cost reduction through caching\n