feat: OpenVINO CPU MetricGAN mask; 100 ms hop

Per-worker OpenVINO CPU runs the MetricGAN BLSTM mask; SpeechBrain
keeps STFT/ISTFT. ENHANCE_INFER=auto with torch fallback. Hop is
0.1 s after isolated last_ms ~6 ms (model+ov+beamform).

tests/test_audio_cleanup.py

@@ -94,7 +94,8 @@ def test_process_records_last_dt_ms():
 def test_stereo_beamform_then_model_preserves_hop():
     c = AudioCleaner(
         mode="force", beamform="force", chunk_s=0.5, hop_s=0.25)
-    assert c.backend == "model+beamform"
+    assert "model" in c.backend
+    assert "beamform" in c.backend
     hop = 4000
     t = np.arange(hop)
     ch0 = (np.sin(2 * np.pi * 440 * t / 16000) * 8000).astype(np.int16)
@@ -104,3 +105,19 @@ def test_stereo_beamform_then_model_preserves_hop():
         out = c.process(stereo)
         assert out.dtype == np.int16
         assert out.shape == (hop,)
+
+
+def test_openvino_cpu_infer_loads_and_preserves_hop():
+    c = AudioCleaner(
+        mode="force", beamform="off", enhance_infer="openvino",
+        chunk_s=0.5, hop_s=0.25)
+    assert c.infer_backend == "openvino"
+    assert c._ov_compiled is not None
+    assert "ov" in c.backend
+    hop = 4000
+    pcm = (np.sin(2 * np.pi * 440 * np.arange(hop) / 16000) * 8000).astype(
+        np.int16)
+    for _ in range(4):
+        out = c.process(pcm)
+        assert out.dtype == np.int16
+        assert out.shape == (hop,)

tests/test_config.py

@@ -13,3 +13,7 @@ def test_default_beamform_mode_auto():
 
 def test_torch_num_threads_default_one():
     assert Config().torch_num_threads == 1
+
+
+def test_enhance_infer_default_auto():
+    assert Config().enhance_infer == "auto"