Confirm that the value used as the swap thrashing threshold in 
swap_thrashing_monitor_delegate_win.h is adequate, it's not clear if this number 
should be higher on more powerful machines. The assumption (based on local 
experiments) is that we rarely observe a continuous page-fault
activity on a high-end system, instead we see some quick bursts (one sample will 
be really high but the next ones will be low). On low-end system the observations
that the page-fault rate will not be too high but will last longer, so this low
threshold seems adequate and works in some local tests but this should be 
adjusted once we get more metrics.