JIT breaks on feather3

[Mark Glines]

One important point: the mprotect() manpage on my machine claims that 
POSIX says mprotect()'s behavior is undefined when applied to a memory 
region that was not obtained with mmap().  So that means we shouldn't be 
using the heap for this, to begin with.  So the next step is to convert 
mem_alloc_executable/mem_free_executable to use mmap instead of 
posix_memalloc (which backends to malloc).

I've done some work on this, but don't have it working yet.  I have 
created trac #18 to track this issue: 
https://trac.parrot.org/parrot/ticket/18

Mark