Add a opnorm function

[MaxenceGollier]

Would it be possible to add something similar to LinearAlgebra.opnorm for the spectral norm of a matrix A ?
The function would essentially compute the largest singular value of A, I suppose we can do something with the iterative methods from this repo ?
I might try to open a PR if this sounds possible/good ?

[amontoison]

I know some iterative algorithms to compute the maximum eigenvalue of a symmetric matrix.
You can use it to compute the maximum singular value if you apply it on A'A.

-> https://en.m.wikipedia.org/wiki/Rayleigh_quotient_iteration

Note you can already do what you want with PROPACK.jl but only on CPU and Float32 / Float64.

Pull requests are welcomed for a generic / GPU version.

[dpo]

That would be a major enterprise, and we don’t want a half-baked implementation.