MPBArray: constructor

MPBArray(AH::AbstractArray{Float64,2}; TF=Float32, TR=nothing)

An MPBArray stores the high precision matrix, the low precision factorization and a few other things BiCGSTAB needs. If the high precision matrix is double, the low precision is single by default. Half is an option which you get with TF=Float16.

MPBArray(AH::AbstractArray{Float32,2}; TF=Float16, TR=nothing)

An MPBArray stores the high precision matrix, the low precision factorization and a few other things BiCGSTAB needs. Since High precision is single, low is half. I'm leaving the kwarg for TF in there because it makes is easier to cut/paste calls to MPBArray different precisions into a CI loop.