Add Qwen3-VL-4B-Instruct to tool call parser map

[cdreetz]

Adds Qwen/Qwen3-VL-4B-Instruct to MODEL_TOOL_CALL_PARSER with hermes parser.

Without this, tool_call_parser = "auto" resolves to None for Qwen3-VL models, causing vLLM to reject tool call requests with 400 errors.

---

Note

Low Risk
Low risk: a single mapping entry change that only affects tool-calling behavior when tool_call_parser="auto" is used for this specific model.

Overview
Adds Qwen/Qwen3-VL-4B-Instruct to MODEL_TOOL_CALL_PARSER with the hermes tool-call parser so resolve_tool_call_parser() can auto-select a parser for this model (avoiding None and resulting tool-call request failures when tool_call_parser is set to "auto").

^{Written by Cursor Bugbot for commit 26ef5e2. This will update automatically on new commits. Configure here.}

[hallerite]

let's add all Qwen3-VL models while we are at it