SVD implementation

It's great that LU / QR / Hessenberg etc was added, but I think SVD would be golden to have as an additional linalg component. A lot of the algorithms from Python bases themself on the SVD decomposition, especially the compact variant.

Is this on the agenda to be added?

Start by reviewing the existing LU, QR, and Hessenberg linear-algebra components to understand how a decomposition is structured in Multik. Then clarify the required SVD and compact-variant scope, including expected APIs and validation; the issue does not name implementation files or tests, so completion criteria are not yet defined.

Written by the indexing model from the issue text.