### Check for existing issues

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

### What happened?

I am testing the LiteLLM `rerank` endpoint against a `hosted_vllm` deployment with `input_cost_per_query` configured. The request succeeds and returns ranked results, but the recorded cost is always zero

This is similar to #7258, where Bedrock rerank cost tracking was fixed by including one `search_unit` in the normalized response. The Hosted vLLM adapter currently records only `total_tokens`, while `BaseRerankConfig.calculate_rerank_cost` calculates per-query cost from `input_cost_per_query * search_units`

I expected every successful uncached Hosted vLLM rerank request to record `search_units=1` and use the configured per-query price

### Steps to Reproduce

1. Configure a `hosted_vllm` rerank deployment with `input_cost_per_query` 2. Send a successful request to `/v1/rerank` 3. Check the request cost or call `calculate_rerank_cost` with the normalized `billed_units` 4. Observe that the response contains `total_tokens` but no `search_units`, so the calculated cost is `0.0`

Minimal reproduction:

```python from litellm.llms.hosted_vllm.rerank…