[Bugfix][Qwen3-TTS] Preserve original model ID in omni_snapshot_download

[linyueqian]

Purpose

Fix the root cause of #1317: omni_snapshot_download() (changed by PR #1213) now returns the resolved HF cache path (e.g. /root/.cache/huggingface/hub/models--Qwen--Qwen3-TTS-12Hz-1.7B-CustomVoice/snapshots/0c0e...) instead of the original model ID. This breaks qwen3_tts.py which parses model_config.model to extract the task type, and could silently break any other code that expects HuggingFace model ID semantics.

Fix: Call download_weights_from_hf_specific() for its pre-caching side effect but always return the original model_id. Also keep the defensive .split("/")[0] in qwen3_tts.py as a safeguard.

Test Plan

python -m pytest tests/e2e/online_serving/test_qwen3_tts.py -v

Test Result

Waiting for @Gaohan123 to re-enable the Qwen3-TTS e2e test in CI (disabled in #1306 due to this bug).

---

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.

[Gaohan123]

LGTM, Thanks!