Add specification for computing singular values using singular value decomposition (linalg: svdvals)

[kgryte]

This PR

• adds a specification for computing singular values using singular value decomposition.

Notes

• This API deviates from current array libraries which overload the svd function modeled after NumPy via a compute_uv keyword. In that API, when compute_uv is False, only the singular values are returned. As discussed in a thread on gh-114 and in a prior weekly meeting, in order to ensure API consistency with eig and eigvals and to limit polymorphism, splitting svd into svd and svdvals was deemed desirable. The former API is equivalent to compute_uv=True, while the latter is equivalent to compute_uv=False.

[kgryte]

This PR has been discussed and approved during meetings. Will merge, and we can submit follow-up PRs should any issues/concerns arise.