This documentation provides a discussion of the tradeoffs of different Power Flow methods in PowerModels,
https://lanl-ansi.github.io/PowerModels.jl/stable/power-flow/
I suggest starting there. Most likely, switching from solve_pf to compute_ac_pf will give you the bulk of the runtime benefits for fast solving AC Power Flows. Going deeper into memory allocation savings in compute_ac_pf, as per the code you reference, I expect is no more than a 2x runtime improvement over the basic implementation that is provided.