Arm backend: Fix bmm quantization bug (pytorch#19798)

bmm nodes are now forwarded to ArmPass in stead of ExportPass.

This fixes an issue where _call_quantized_bmm_without_fake_kernel() 
does not get called, leading to dtype mismatch error


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

Signed-off-by: Christoffer J.L <christoffer.johanssonlundqvist@arm.com>

Author: Christoffer-JL

backends/arm/_passes/replace_scalar_with_tensor_pass.py

@@ -126,4 +126,4 @@ def call_operator(self, op, args, kwargs, meta):
             return super().call_operator(op, args, kwargs, meta)
         else:
             # Do not handle; forward unchanged.
-            return ExportPass.call_operator(self, op, args, kwargs, meta)
+            return ArmPass.call_operator(self, op, args, kwargs, meta)