Adapt sym quantizer to ET (#18870)

Summary:

# Summary
This diff includes a symmetric 8 bit quantizer in ET.

---
#hthtemplate

Reviewed By: hsharma35

Differential Revision: D91777784

Author: Marco Giordano

backends/cadence/aot/quantizer/quantizer.py

@@ -97,6 +97,15 @@
     observer_or_fake_quant_ctr=MinMaxObserver,
 )
 
+wgt_qspec_sym8s_127 = QuantizationSpec(
+    dtype=torch.int8,
+    quant_min=-127,
+    quant_max=127,
+    qscheme=torch.per_tensor_symmetric,
+    is_dynamic=False,
+    observer_or_fake_quant_ctr=MinMaxObserver,
+)
+
 bias_qspec: Optional[QuantizationSpec] = None
 
 qconfig_A8W8 = QuantizationConfig(
@@ -161,11 +170,11 @@
     None,
 )
 
-qconfig_A32W8sym = QuantizationConfig(
+qconfig_A32W8sym_127 = QuantizationConfig(
     input_activation=None,
     output_activation=None,
-    weight=wgt_qspec_sym8s,
-    bias=wgt_qspec_sym8s,
+    weight=wgt_qspec_sym8s_127,
+    bias=wgt_qspec_sym8s_127,
 )
 
 
@@ -417,13 +426,13 @@ class CadenceW8A32MixedQuantizer(CadenceQuantizer):
     def __init__(self) -> None:
         quantizers = []
         quantizers.append(
-            CadenceAtenQuantizer(MixedW8A32LinearPattern(), qconfig_A32W8sym)
+            CadenceAtenQuantizer(MixedW8A32LinearPattern(), qconfig_A32W8sym_127)
         )
         quantizers.append(
-            CadenceAtenQuantizer(MixedW8A32ConvPattern(), qconfig_A32W8sym)
+            CadenceAtenQuantizer(MixedW8A32ConvPattern(), qconfig_A32W8sym_127)
         )
         quantizers.append(
-            CadenceAtenQuantizer(MixedW8A32GruPattern(), qconfig_A32W8sym)
+            CadenceAtenQuantizer(MixedW8A32GruPattern(), qconfig_A32W8sym_127)
         )
         super().__init__(quantizers)