fix(onnx-asr): исправлены баги интеграции после smoke-теста

- Зачем:
  - при smoke-тесте выявлены расхождения между onnx-asr API и мок-тестами.
- Что:
  - убран deprecated cpu_preprocessing из load_model.
  - sample_rate передан как keyword-аргумент в recognize().
  - исправлены имена полей SegmentResult: start_ts→start, end_ts→end.
  - добавлен device "onnx" в DEVICE_DEFAULTS и _VALID_DEVICES.
- Проверка:
  - uv run pytest -q (172 passed).
  - uv run transcribe file.mp4 --device onnx --model gigaam-v3.

tests/test_onnx_asr.py

@@ -10,9 +10,9 @@ from local_transcriber.types import Segment, TranscribeResult
 class FakeVadSegment:
     """Mimics onnx-asr SegmentResult."""
 
-    def __init__(self, start_ts, end_ts, text):
-        self.start_ts = start_ts
-        self.end_ts = end_ts
+    def __init__(self, start, end, text):
+        self.start = start
+        self.end = end
         self.text = text
 
 
@@ -40,10 +40,9 @@ class TestCreateModel:
         calls = []
 
         def fake_load_model(model=None, path=None, quantization=None,
-                           cpu_preprocessing=None, **kwargs):
+                           **kwargs):
             calls.append({
                 "model": model, "path": path, "quantization": quantization,
-                "cpu_preprocessing": cpu_preprocessing,
             })
             return FakeAsrAdapter()
 
@@ -59,7 +58,6 @@ class TestCreateModel:
 
         assert len(calls) == 1
         assert calls[0]["quantization"] == "int8"
-        assert calls[0]["cpu_preprocessing"] is True
         assert model is not None
 
     def test_loads_silero_vad(self, monkeypatch):