more robust beyond-reservation reachability handling

_From [bruen...@google.com](https://code.google.com/u/109494838902877177630/) on March 25, 2013 12:03:48_

split from issue #774 today we reserve 128MB, and if we run out, we do ensure new allocs are reachable by adding each reservation to the must-reach range.

first, we should only do that for vmcode, after issue #1132 .

second, we may want to tweak things to reduce the chance of OOM:
- up the reservation to 2GB?
- be more aggressive about resets when we near the 2GB limit?

_Original issue: http://code.google.com/p/dynamorio/issues/detail?id=1133_