Use Box in SVM and remove lifetimes

[morenol]

Fixes #

• Remove lifetimes in Kernel trait.
• Use box instead of & for Kernel objects n SVM structs.

Checklist

• My branch is up-to-date with development branch.
• Everything works and tested on latest stable Rust.
• Coverage and Linting have been applied

Current behaviour

New expected behaviour

Change logs

[Mec-iS]

i like the idea of making the kernels name with static lifetime but we cannot use Box for the kernel as it ruins the API call. For an example of how kernels are used see the Jupyter notebooks:

let y_hat: Vec<f64> = SVC::fit(
    &x, &y,
    &SVCParameters::default()
        .with_c(1.0)
        .with_kernel(&Kernels::rbf()
            .with_gamma(0.7)),
     ) .and_then(
        |lr| lr.predict(&x)
) .unwrap(); 

how the kernels are used in tests is irrelevant. we don't want the end-users to allocate to the heap manually.
Kernels are intended to be used as references in parameters and it is made possible by them being object-safe (dyn...). As far as I understood, for object-safe trait using & or using box is the same as they are dynamically allocated at runtime, see https://www.reddit.com/r/rust/comments/kw8p5v/what_does_it_mean_to_be_an_objectsafe_trait/ or "Advanced Traits" in the Rust book.

[Mec-iS]

about implementing an enum for kernels, see #221

[Mec-iS]

be careful. I agreed in using 'static for the strings used by kernels' names, but if you implement the whole Kernel as 'static itself it means that you are keeping in memory every kernel instantiated for the entire length of the program.

My idea was to make the string names static, then implement an enumerator to convert string->kernel object, see #221 there is a library that allows doing string to object instantiation but I am not sure that it is viable.

[Mec-iS]

@morenol Ok then, please keep the &dyn Kernel syntax for type declarations as it is the preferred one since Rust 2021.
The rest is OK.

Good one!