Fix incorrect assumption when parsing tool_calls[].function.arguments

[LxYuan0420]

This PR is:

• to fix the incorrect assumption that tool_calls[].function.arguments is always a JSON string
• to parse function.arguments with json.loads() only when it is a string, and leave dict untouched

Reproduce command:

$ mlx_lm.server \
  --model mlx-community/functiongemma-270m-it-4bit \
  --port 8081


$ curl -i http://127.0.0.1:8081/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"default_model","messages":[{"role":"user","content":"what is 2+3?"},{"role":"assistant","content":null,"tool_calls":
  [{"type":"function","id":"call_123","function":{"name":"add","arguments":{"a":2,"b":3}}}]},
  {"role":"tool","name":"add","tool_call_id":"call_123","content":"5"}],"max_tokens":16}'

Before fix:

HTTP/1.0 404 Not Found
{"error": "the JSON object must be str, bytes or bytearray, not dict"}

After fix:

HTTP/1.0 200 OK
chat.completion response

TL;DR: If function.arguments is a string, parse it; if it’s already a dict, leave it untouched.

• ☑️ I understand it is strictly prohibited to use AI to write PR description