fix:Enhance ASR Service Stability and API Performance

Summary

Minor fixes to ensure the backend runs correctly after splitting into separate frontend and backend repositories.

Changes

app/main.py

• Added eager model pre-loading in the startup_event handler.
• Previously the ASR model was loaded slowly on the first request, causing a noticeable cold-start delay for the first user. Now load_model() is called at application startup, wrapped in a try/except so startup failures are logged but do not crash the server.

app/asr_service.py

• Cleaned up inline comments around the pipe() transcription call for better readability (no functional change).

Impact

• First-request latency is eliminated since the model is warm when the server starts.
• No breaking changes; all existing endpoints remain unchanged.

Testing

• Verified the server starts without errors after the repo split.
• Confirmed model loads during startup log output.