Add streaming multi-talker Parakeet RNN-T ASR model

Add NVIDIA NeMo `EncDecMultiTalkerRNNTBPEModel`
(multitalker-parakeet-streaming-0.6b-v1) as its own model type,
exported as three ONNX graphs matching the nemotron-speech-streaming
cache-aware streaming deployment contract:

- encoder: FastConformer with speaker-kernel injection and streaming
  channel/time caches (causal dw_striding subsampling, bias-free
  linears, LayerNorm conv, chunked-limited attention, rel-pos KV cache).
- decoder: RNN-T LSTM prediction network (2-layer ONNX LSTM with
  PyTorch->ONNX gate reordering).
- joint: RNN-T joint network emitting per-(time, target) logits.

Adds MultiTalkerRNNTTask, the ParakeetMultiTalkerModel/Config classes
with a .nemo loader (build_parakeet_multitalker), unit tests (graph
build + ORT smoke run of all three graphs), and an integration test
verifying streaming parity against the NeMo PyTorch reference
(encoder 3.9e-7, decoder 1.1e-6, joint argmax match).

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: Mason Corey <masoncorey@microsoft.com>

Author: themason2011

src/mobius/models/__init__.py

@@ -14,6 +14,7 @@
     "Blip2Model",
     "BloomCausalLMModel",
     "build_sortformer",
+    "build_parakeet_multitalker",
     "CLIPVisionModel",
     "SigLIPVisionModel",
     "CTRLCausalLMModel",
@@ -124,6 +125,8 @@
     "SmolLM3CausalLMModel",
     "SortformerConfig",
     "SortformerDiarizationModel",
+    "ParakeetMultiTalkerConfig",
+    "ParakeetMultiTalkerModel",
     "StarCoder2CausalLMModel",
     "T2IAdapterModel",
     "T5ForConditionalGeneration",
@@ -258,6 +261,11 @@
 )
 from mobius.models.sensevoice_small import SenseVoiceSmallModel
 from mobius.models.smollm import SmolLM3CausalLMModel
+from mobius.models.parakeet_multitalker import (
+    ParakeetMultiTalkerConfig,
+    ParakeetMultiTalkerModel,
+    build_parakeet_multitalker,
+)
 from mobius.models.sortformer import (
     SortformerConfig,
     SortformerDiarizationModel,