[Docs] Clarify VLM LoRA serving scope

[zcxGGmu]

Motivation

Issue #3981 asks whether LMDeploy supports serving multiple LoRA adapters for multimodal models and notes that vision-side LoRA did not work in their test.

Current InternVL model wrappers route LoRA loading through the language model when that loader is available, so the existing generic LoRA serving docs should clarify the VLM scope instead of implying that arbitrary visual-encoder adapters are covered.

Refs #3981

Modification

• Clarify in English and Chinese LoRA serving docs that the generic serving path is intended for language-model adapters on multimodal models.
• Mention that vision-encoder LoRA adapters are not covered by the generic serving path and should be merged or validated through a model-specific path before serving.

BC-breaking (Optional)

No. Documentation only.

Use cases (Optional)

N/A

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
   • PYTHONPATH= python -m pre_commit run --files docs/en/llm/api_server_lora.md docs/zh_cn/llm/api_server_lora.md
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
   • Documentation-only change; verified by grep for the new VLM LoRA scope text.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
   • N/A
4. The documentation has been modified accordingly, like docstring or example tutorials.
   • Updated docs/en/llm/api_server_lora.md and docs/zh_cn/llm/api_server_lora.md.