[Doc] Update max_tokens to max_completion_tokens in all docs (#6248)

### What this PR does / why we need it?

Fix:

```
DeprecationWarning: max_tokens is deprecated in favor of the max_completion_tokens field.
```

- vLLM version: v0.14.1
- vLLM main:
d68209402d

Signed-off-by: shen-shanshan <467638484@qq.com>

docs/source/quick_start.md

@@ -168,7 +168,7 @@ curl http://localhost:8000/v1/completions \
     -d '{
         "model": "Qwen/Qwen2.5-0.5B-Instruct",
         "prompt": "Beijing is a",
-        "max_tokens": 5,
+        "max_completion_tokens": 5,
         "temperature": 0
     }' | python3 -m json.tool
 ```