Use tenant-scoped dubbed-audio paths with traversal-safe resolution

[AJaySi]

Motivation

• Replace the hard-coded global DUBBED_AUDIO_DIR with the same tenant-first media layout used for podcast audio/video so outputs live under each user workspace.
• Require user-scoped directory lookup and traversal-safe resolution to prevent path traversal when serving dubbed audio or cloned voice files.

Description

• Added _get_dubbed_audio_dir(user_id, *, ensure_exists=False) to resolve per-tenant dubbed_audio directories under the podcast audio workspace and create them when required.
• Added _resolve_dubbed_audio_file(filename, user_id) which searches tenant directories first, falls back to the legacy data/media/dubbed_audio path, and enforces traversal-safe checks before returning a file path.
• Replaced usages of the global DUBBED_AUDIO_DIR by initializing AudioDubbingService(output_dir=...) with the tenant-specific directory and by serving files via _resolve_dubbed_audio_file(...) in /dub/audio/{filename} and /dub/voices/audio/{filename} handlers.
• Preserved backward compatibility by including the legacy data/media/dubbed_audio directory as a final fallback read location.

Testing

• Compiled the modified module with python -m py_compile backend/api/podcast/handlers/dubbing.py, which succeeded.

---

Codex Task