Cortex_m backend: Simplify add + linear fusion passes

Reuses the FoldAndAnnotateQParamsPass from the Arm backend
to greatly simplify the logic for fusing the ops.

Additionally updates the linear kernel to be numerically correct
and computes the kernel_sum aot in the quantized_linear_fusion pass.
Note that since this replaces the bias node it typically causes no
extra memory useage.

Updates the Linear tests to mirror this, including removing the
various matmul tests. Since the linear is handled as a separate op
rather than a particular type of matmul these tests are not related
anymore.

Removes unnecessary stub definitions in operatos.py, operators.yaml and
op_quantized_linear.cpp

Leaving a few TODO:s since the patch is large already.

Signed-off-by: Adrian Lundell <[email protected]>
Change-Id: I194228ee3ae4b64a92f3f818afb2e045cc3acf91

Author: AdrianLundell

backends/cortex_m/ops/cmsis_scratch_buffer_context.h

@@ -22,6 +22,7 @@ using Tensor = torch::executor::Tensor;
 using ScalarType = executorch::aten::ScalarType;
 using Scalar = torch::executor::Scalar;
 using Error = executorch::runtime::Error;
+using IntArrayRef = executorch::aten::ArrayRef<int64_t>;
 
 // From arm_nn_math_types.h
 #define ARM_NN_Q31_MAX ((int32_t)(0x7FFFFFFFL))