Move fault page allocation into its own routine.

[jeff]

Summary

This change has a small functional difference. oom now counts the number of successive page allocation failures and is reset when there is no failure. It used to reset when there was any restart condition unrelated to oom. I don't have strong feelings about it but I think this is a better test. Blocking on a backing busy lock should not clear oom IMO. If we want the old behavior I need a different return code for OOM.