- [function] computation of input dimensions now happens in 'MatMulImpl_cpu::forward()'. It handle indexes and sends the right pointers to its computation kernel
- [function] 'getCPUPtr()' new parameter: offset
- [function] reduce MatMul kernel function to simple matrices multiplication

- [function] computation of input dimensions now happens in 'MatMulImpl_cpu::forward()'. It handle indexes and sends the right pointers to its computation kernel
- [function] 'getCPUPtr()' new parameter: offset
- [function] reduce MatMul kernel function to simple matrices multiplication

[fix/rawPtr][MIN][DSGN] encapsulating raw pointer access into an helper function in order to make it easier to change implementation related to aidge_core#21 and related to aidge_core!41 (which will get easier to merge threafter)

- adapt to core changes with Operator not refering to Tensors anymore
- remove assertions already performed in abstract operators constructions
- fix missing 'template' keyword prior to dependent template name 'dims'