Matmul rework

The behavior of the reworked MatMul on the arguments in the following way.
- If both arguments are 2-D they are multiplied like conventional matrices.
- If either argument is N-D, N > 2, it is treated as a stack of matrices residing in the last two indexes.
Exp: For the MatMul of tensor A of shape (3,3,2) with tensor B of shape (3,2,4), A is a stack of three 2D matrices each of shape (3,2), and B is a stack of 3 2D matrices, each of shape (2,4).

include/aidge/backend/cpu/operator/MatMulImpl.hpp

@@ -27,12 +27,12 @@ namespace Aidge {
 
 // compute kernel registry for forward and backward
 class MatMulImplForward_cpu
-    : public Registrable<MatMulImplForward_cpu, std::tuple<DataType, DataType, DataType>,
-                         void(const MatMul_Op::Attrs &, const DimSize_t, const DimSize_t,
+    : public Registrable<MatMulImplForward_cpu, std::tuple<DataType, DataType>,
+                         void(const std::vector<DimSize_t>&, const std::vector<DimSize_t>&,
                               const void *, const void *, void *)> {};
 class MatMulImplBackward_cpu
-    : public Registrable<MatMulImplBackward_cpu, std::tuple<DataType, DataType, DataType>,
-                         void(const MatMul_Op::Attrs &, const DimSize_t, const DimSize_t,
+    : public Registrable<MatMulImplBackward_cpu, std::tuple<DataType, DataType>,
+                         void(const std::vector<DimSize_t>&, const std::vector<DimSize_t>&,
                               const void *, const void *, void *)> {};
 
 class MatMulImpl_cpu : public OperatorImpl {