fix(kokoro): trim edges only, keep inter-segment gaps

Registering Kokoro with needs_trim=True routed its output through the
generic trim_tts_output, whose 1s internal-silence cut was tuned for
Chatterbox hallucinations. KPipeline synthesizes newline- and
token-limit-separated segments independently, and the ~0.3s lead plus
~0.7s tail pads at each boundary add up to 1.2s of silence, so with
af_sarah a four-paragraph script came back as its first paragraph only
(10.9s -> 1.8s) and a 51s text lost half its segments.

Drop the needs_trim flag, keep the in-backend trim, and give
trim_tts_output a max_internal_silence_ms=None mode that only trims the
leading and trailing pads. Tests updated for the new behaviour, with a
two-segment fake pipeline and an explicit internal-gap case.

backend/backends/__init__.py

@@ -369,7 +369,6 @@ def _get_non_qwen_tts_configs() -> list[ModelConfig]:
             engine="kokoro",
             hf_repo_id="hexgrad/Kokoro-82M",
             size_mb=350,
-            needs_trim=True,
             languages=["en", "es", "fr", "hi", "it", "pt", "ja", "zh"],
         ),
     ]

backend/backends/kokoro_backend.py

@@ -291,7 +291,12 @@ class KokoroTTSBackend:
             audio = np.concatenate(audio_chunks).astype(np.float32)
             from ..utils.audio import trim_tts_output
 
-            audio = trim_tts_output(audio, sample_rate=KOKORO_SAMPLE_RATE)
+            # Edge-only trim: Kokoro pads ~0.3s before and ~0.7s after speech.
+            # The internal-gap cut is disabled because KPipeline synthesizes
+            # newline/token-limit segments independently and the pads at each
+            # segment boundary add up to >1s of silence; with the default cut
+            # everything after the first segment would be dropped.
+            audio = trim_tts_output(audio, sample_rate=KOKORO_SAMPLE_RATE, max_internal_silence_ms=None)
             return audio, KOKORO_SAMPLE_RATE
 
         return await asyncio.to_thread(_generate_sync)