fix(tpu): fix LoRA and V1 engine crashes

[chiajunglien]

Key Changes

• Fixed LoRA Device Retrieval (utils.py): Added support for nn.ParameterList in _get_lora_device. This prevents AttributeError when weights are stored in containers, which is common in TPU and MoE implementations.

The following command was used to verify the fixes on TPU v6e:
MODEL_IMPL_TYPE=vllm TPU_BACKEND_TYPE=jax python3 examples/offline_inference/multilora_inference.py

Purpose

Test Plan

Test Result

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

BEFORE SUBMITTING, PLEASE READ https://docs.vllm.ai/en/latest/contributing (anything written below this line will be removed by GitHub Actions)