### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

When setting `model_max_budget` on a customer (via separate `/budget/new` with `model_max_budget` + linking to customer), the per-model budget limit is correctly stored in the database and visible via `/customer/info`, but it is **not enforced** at runtime.

I verified this on LiteLLM 1.90.0: set per-model budget `$0.00001` for model `google/gemini-2.5-flash-lite`, made 10 consecutive requests each costing `$0.0001768` — **all passed**, none were blocked.

**Expected:** Requests should be blocked once the per-model spend exceeds the configured per-model budget.

### Steps to Reproduce

```python import httpx, time

client = httpx.Client(base_url="https://your-proxy.com") auth = {"Authorization": "Bearer sk-master-key", "Content-Type": "application/json"}

user_id = "test-model-budget"

# 1. Create a budget with model_max_budget set very low budget_resp = client.post("/budget/new", headers=auth, json={ "budget_duration": "1d", "model_max_budget": { "google/gemini-2.5-flash-lite": { "max_budget": 1e-05, "bu…