addition transpose 2D

[chachaleo]

Pull Request type

• Bugfix
• Feature
• Code style update (formatting, renaming)
• Refactoring (no functional changes, no API changes)
• Build-related changes
• Documentation content changes
• Other (please describe):

What is the current behavior?

In the implementation of tensor.transpose in https://github.com/gizatechxyz/orion/blob/main/src/operators/tensor/linalg/transpose.cairo, the implementation is very general and support any permutations. However, one of the most general use case, matrix transposition, could be done in a more optimized way. As it is used a lot in matrix theory and in linear algebra it can be important to have a more optimized implementation.

Issue Number: #394

What is the new behavior?

If the dimension of the input tensor is 2, the function transpose calls a sub-function transpose2D that computes the transposition of a 2D tensor. The 2D tensor transposition is a trivial permutation of axes and the corresponding retrieval of the element is therefore more straightforward.
It directly gets the data from the input and avoids the creation of temporary arrays to support the more general permutation of the ND tensor problem.

output_data.append(*(*self.data)[i * m + j]);

With this implementation, transposition of 2D tensor is consumes about one order of magnitude less of gas.

[raphaelDkhn]

LGTM! Thanks!