Rework memory handling in ProcessCore

This reworks memory handling in ProcessCore, turning the runnable reading smaps_rollup into a long-lived single thread that is fairly restrictive in update rate. It also keeps track of two sources of memory information, one precise from smaps_rollup and one imprecise from statm. Since the precise data is expensive to retrieve, we calculate the delta between precise and imprecise data and use that as a correction factor on the imprecise data whenever the imprecise data is updated.