We should track the total time spent and the time spent in TCMalloc so we can understand where time is going globally.
I think we should shard them by CurrentCore() to avoid contention and get more granular metrics. We want a timer for the amount of time spent in SystemAllocator. We probably also want counters for how many times we go down each code path in BufferAllocator::AllocateInternal() (i.e. getting a hit immediately in the local area, evicting a clean page, etc down to doing a full locked scavenge).