Add eigen for general purpose computation of eigenvectors/eigenvalues

[GVigne]

Is your feature request related to a problem? Please describe.
I would like to be able to compute the eigenvectors and eigenvalues a matrix which can either hold floats or complex. I didn't see a generic way to do this (like LinearAlgebra's eigen) and instead have to manually check the type before calling the appropriate functions, namely syevd!and heevd! (the matrix is always Hermitian/symmetric).

Describe the solution you'd like
A high level function taking a matrix A and returning, if possible, the eigenvectors and eigenvalues of A. This function may throw if no appropriate function was found (I think there aren't a lot of solvers for non-Hermitian complex matrixces in cuSolver).

Describe alternatives you've considered
Right now, I am doing something like

if eltype(A) <: Complex
        vals, vects = CUDA.CUSOLVER.heevd!('V','U',A)
    else
        vals, vects = CUDA.CUSOLVER.syevd!('V','U',A)
    end

This is the type of thing I imagined eigen could do.

Additional context
Maybe there is a very good reason as why this wasn't implemented, but I couldn't find it in the docs or the code. If eigen is not to be implemented, maybe we could add a little paragraph somewhere to explain this?

[maleadt]

We seem to have this now:

CUDA.jl/lib/cusolver/linalg.jl

Lines 115 to 125 in 2a3cec9

	function LinearAlgebra.eigen(A::Symmetric{T,<:CuMatrix}) where {T<:BlasReal}
	A2 = copy(A.data)
	Eigen(syevd!('V', 'U', A2)...)
	end
	function LinearAlgebra.eigen(A::Hermitian{T,<:CuMatrix}) where {T<:BlasComplex}
	A2 = copy(A.data)
	Eigen(heevd!('V', 'U', A2)...)
	end
	function LinearAlgebra.eigen(A::Hermitian{T,<:CuMatrix}) where {T<:BlasReal}
	eigen(Symmetric(A))
	end