Memory-efficient Optimization of Gyrokinetic Particle-to-Grid Interpolation for Multicore Processors