### Check for existing issues

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

### What happened?

`afile_content` calls (retrieving file content — a files/storage API call, not an LLM completion) are billed a `response_cost`, even though there is no real token usage to bill. The cost consistently equals **exactly 3 × the model's `input_cost_per_token`** — e.g. for `vertex_ai/gemini-3-flash-preview` (`input_cost_per_token = 5e-07`), the recorded cost is `1.5e-6`, which is `3 × 5e-07`. This pattern appears across all `afile_content` calls where a response cost is recorded, regardless of the actual file size or content.

**Root cause:** `afile_content` is not a first-class, documented call type in `cost_calculator.py`'s Literal list and has no dedicated pricing rule. However, `afile_content` calls go through the same `@client` logging path as chat-style completions, and `HttpxBinaryResponseContent` is treated as a normal logged response — so the generic `response_cost` / `completion_cost` path still runs. Because `function_setup` never fills in real messages for this call type, it falls back to the placeholder string `"default-message-value"`,…