sparse by dense parallel products #298

[aujxn]

Starts on #298. Using ndarray parallel iterators this is the naive conversion of the sparse by dense products to a multithreaded implementation. Ideally, there is a rayon into_par_iter implemenation for the existing CsMat::outer_iterator for some of these to be more optimal. In the current implementation, some sparse matrix by dense vector products are still going to be single threaded.

[mulimoen]

Lets bump MSRV and see if the tests pass

[mulimoen]

We will also need to run with --features multi-thread on CI to test this implementation