Arm backend: Add missing BasePipeline.set_quantization_calibration (pytorch#23265)

pytorch#23182 made the NSS and NFRU model tests call
pipeline.set_quantization_calibration(), but never added the method, so
the real-data TOSA INT tests fail with AttributeError. Add it to
BasePipeline, configuring the quantize stage the same way the NSS test
did directly before pytorch#23182.

Authored with Claude Code.

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils
@Sebastian-Larsson @robell @rascani

Author: psiddh

backends/arm/test/tester/test_pipeline.py

@@ -11,6 +11,7 @@
     Callable,
     Dict,
     Generic,
+    Iterable,
     List,
     Optional,
     Sequence,
@@ -246,6 +247,20 @@ def quantizer(self) -> TOSAQuantizer:
                 f"First argument of quantize stage was {type(quantize_stage).__name__}, not Quantize as expected."
             )
 
+    def set_quantization_calibration(
+        self,
+        calibration_samples: Iterable[Any],
+        dynamic_shapes: Optional[Tuple[Any, ...]] = None,
+    ):
+        """Calibrates the quantize stage with the given samples instead of the
+        test data.
+        """
+        quantize_stage = self._stages[self.find_pos("quantize")].args[0]
+        quantize_stage.calibration_samples = calibration_samples
+        if dynamic_shapes is not None:
+            quantize_stage.dynamic_shapes = dynamic_shapes
+        return self
+
     def pop_stage(self, identifier: int | str):
         """Removes and returns the stage at postion pos."""
         if isinstance(identifier, int):