[request] Update nalgebra dependency

[saona-raimundo]

I was using the MultivariateNormal. Specifically, the continuous trait.

In my case, I had the data in ndarray, so I passed it to nalgebra. But the nalgebra has a new version of 1d arrays.

For reference, this is the extract from rustc

                              gaussians[i].pdf(&nalgebra::base::DVector::from(point.to_vec())))
    |                                                --- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `&Matrix<f64, Dynamic, ..., ...>`, found `&Matrix<f64, Dyn, Const<1>, ...>`
    |                                                |
    |                                                arguments to this method are incorrect
    |
    = note: `Matrix<f64, Dyn, Const<1>, ...>` and `Matrix<f64, Dynamic, Const<1>, ...>` have similar names, but are actually distinct types
note: `Matrix<f64, Dyn, Const<1>, ...>` is defined in crate `nalgebra`
   --> ...\nalgebra-0.32.3\src\base\matrix.rs:175:1
    |
175 | pub struct Matrix<T, R, C, S> {
    | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
note: `Matrix<f64, Dynamic, Const<1>, ...>` is defined in crate `nalgebra`
   --> ...\nalgebra-0.29.0\src\base\matrix.rs:157:1
    |
157 | pub struct Matrix<T, R, C, S> {
    | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    = note: perhaps two different versions of crate `nalgebra` are being used?

[saona-raimundo]

As a second thought, would it not be better to impl<'a> Continuous<&'a [f64]> for MultivariateNormal?

[Heeten]

I'm running into friction due to the older nalgebra version as well. Just curious what the best way to get traction on this is? It looks like there's already PR #187 and multiple other PRs to upgrade the version.

[saona-raimundo]

The implementation of impl Continuous<Vec<f64>> for MultivariateNormal solves the issue.