fix(backend): keep the trimmed clip when a runaway retry cannot split further; track the S3 tokenizer repo in model status and delete

Review follow-ups: with retries_runaway on for an engine that also has a
trim step, a <=100-char chunk flagged as runaway raised instead of
falling back to the trimmed output that already cuts the silence+noise
tail; generate_chunked now returns the trimmed chunk in that terminal
case. ModelConfig gains aux_hf_repo_ids so /models status only reports
the Chatterbox MLX model as downloaded once mlx-community/S3TokenizerV2
is present too (matching the backend's own cache check) and
DELETE /models/{name} removes that repo as well.

backend/utils/chunked_tts.py

@@ -265,6 +265,15 @@ async def generate_chunked(
 
         if runaway_detector is not None and runaway_detector(chunk_audio, chunk_sr):
             if retry_depth >= MAX_RUNAWAY_RETRIES or len(chunk_text) <= MIN_RUNAWAY_RETRY_CHARS:
+                if trim_fn is not None:
+                    # Engines with a trim step (Chatterbox) already cut the
+                    # silence-then-noise tail; prefer the trimmed clip over
+                    # failing the whole generation when we cannot split further.
+                    logger.warning(
+                        "Unstable TTS output for %d chars could not be retried further; keeping trimmed output",
+                        len(chunk_text),
+                    )
+                    return np.asarray(trim_fn(chunk_audio, chunk_sr), dtype=np.float32), chunk_sr
                 raise RuntimeError(
                     "TTS output remained unstable after retrying smaller text chunks"
                 )