[Feature Request] Whisper STT always outputs Traditional Chinese for Mandarin — please add a Simplified Chinese option

Describe the problem
When using the built-in Whisper speech recognition (STT) in OpenClaw Tray on Windows, Mandarin Chinese speech is consistently transcribed into Traditional Chinese characters, even though simplified Chinese is the expected output for most mainland Chinese users.

Real example (user speaks Mandarin):
"虽然我想选A，但是我有点社恐"
gets transcribed as:
"雖然我有點想選A,但是我是個社恐"

Environment

• OpenClaw Tray on Windows 11 (WSL2 gateway)
• STT model: base (ggml-base.bin)
• STT language setting: zh (also reproduces with auto)

Expected behavior
When the user speaks Mandarin, the transcription should be Simplified Chinese.

Root cause (my guess)
The multilingual whisper.cpp model is trained on a large amount of Traditional Chinese text (e.g. YouTube/subtitle corpora), so small models (tiny/base) strongly favor Traditional Chinese output. Forcing language=zh does not affect the simplified/traditional choice.

Suggested solutions (any one of these would help)

1. Support initial_prompt in the whisper.cpp params, allowing an injected prompt like "以下是简体中文。" — a well-known trick that significantly reduces Traditional Chinese output.
2. Add an optional OpenCC post-processing step (traditional → simplified) behind a settings toggle, e.g. SttSimplifiedChinese: true.
3. Or add a "Simplified Chinese preferred" option to the language list.

Additional notes

• The model list currently only offers tiny/base/small. A larger model (e.g. large-v3-turbo) tends to produce more standardized simplified Chinese but is not available in the list.
• Relevant code locations: SpeechToTextService / AudioPipelineOptions (whisper params) and VoiceSettingsPage.

Thanks for the great work on the Tray!

作为一名中文用户，如果语音识别能支持简体中文，日常使用会更顺畅。非常感谢你们的工作！❤️

Start by reading SpeechToTextService and AudioPipelineOptions to see how Whisper parameters are passed, then inspect VoiceSettingsPage for the existing language and model settings. Decide which proposed approach the project supports, implement the simplified-Chinese behavior or setting, and verify that Mandarin speech produces Simplified Chinese without changing other languages.