drheapstat holds global malloc lock, resulting in DR lock being held when hit was-writable fault!

_From [bruen...@google.com](https://code.google.com/u/109494838902877177630/) on October 27, 2013 15:50:25_

Xref issue #1354 and DRi#143. I'm splitting this from those cases as we have a working solution for Dr. Memory that I'm going to go ahead and commit.

How do we solve this one? I don't think we can mark the replace-malloc DR
lock as dr_mark_safe_to_suspend() as it's acquired in some places where a
regular DR lock is held (e.g., itimer event).

_Original issue: http://code.google.com/p/drmemory/issues/detail?id=1358_