Eventually retire chol_lower and chol_upper

[st--]

Now that JuliaLang/LinearAlgebra.jl#1186 is merged, once it can be used by downstream packages, the chol_lower and chol_upper wrappers in chol.jl can probably be retired. :)

[devmotion]

We could simplify chol_lower and chol_upper but I'm not sure if it's possible to remove them completely: They are also specialized for Matrix and sparse matrices in

PDMats.jl/src/chol.jl

Lines 13 to 25 in 5e7d88e

	# For a dense Matrix, the following allows us to avoid the Adjoint wrapper:
	chol_lower(a::Matrix) = cholesky(Symmetric(a, :L)).L
	# NOTE: Formally, the line above should use Hermitian() instead of Symmetric(),
	# but this currently has an AutoDiff issue in Zygote.jl, and PDMat is
	# type-restricted to be Real, so they are equivalent.
	chol_upper(a::Matrix) = cholesky(Symmetric(a, :U)).U
	if HAVE_CHOLMOD
	CholTypeSparse{T} = SuiteSparse.CHOLMOD.Factor{T}
	# Take into account pivoting!
	chol_lower(cf::CholTypeSparse) = cf.PtL
	chol_upper(cf::CholTypeSparse) = cf.UP