Experimenting with using the physical program counter sample
data (implemented 20070608) together with the "translatable
range" information, to figure out which physical address ranges
would be worth translating to native code (if the number of
samples falling within a range is above a certain threshold).
