[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/tutorials/Qwen3-8B-W4A8.md

@@ -106,7 +106,7 @@ curl http://localhost:8000/v1/completions \
     -d '{
         "model": "qwen3-8b-w4a8",
         "prompt": "what is large language model?",
-        "max_tokens": "128",
+        "max_completion_tokens": "128",
         "top_p": "0.95",
         "top_k": "40",
         "temperature": "0.0"