feat: initial public release v0.1.0 — MLX port of pyannote-speaker-diarization-3.1

Byte-parity with pyannote-PyTorch reference (cosine 0.763718 identical
at 6 decimals on 200 cross-window slot pairs). 2.5x faster than
pyannote-MPS on Apple Silicon native.

Extracted from gitea.tavportal.com/olivier/MLX_CONVERTOR commit 5f9eafa.

tests/unit/test_diar_bilstm.py

@@ -0,0 +1,11 @@
+import mlx.core as mx
+from pyannote_diarization_3_1_mlx._bilstm import BiLSTM4
+
+
+def test_bilstm_output_shape():
+    # input (B, T, hidden_in) — pyannote feeds 60-channel sincnet output
+    # transposed to (B, T, 60). hidden=128, bidirectional → 256 out.
+    net = BiLSTM4(input_size=60, hidden_size=128)
+    x = mx.zeros((1, 589, 60))
+    out = net(x)
+    assert out.shape == (1, 589, 256), f"got {out.shape}"