Improve performance of SVC

[DanielLacina]

In the optimization process, I noticed the implementation involves iterating over each row in the X matrix and then iterating over each value in the row vector. I think we can improve the performance by doing the computations with the abstracted Arrays instead of having to convert them into Rust float vectors. I sent a video to show examples of the current issue: https://github.com/user-attachments/assets/72d00751-8e86-4d6b-8e81-d6e47ef15c2a

[Mec-iS]

Thanks for finding this. That implementation is old, from before the new Array abstraction had been implemented.