[PATCH] [Sanitizer] Update GetMaxVirtualAddress for PowerPC64 Linux

[Will Schmidt]

>>! In D6082#7, @kcc wrote:
> I am always reluctant to support any old OS (and > 2+ years is old) unless there is a specific request and someone is willing to do the support work. 
> My team currently does't test on PowerPC at all. So, the simplest solution (i.e. only support 46-bit and assert at run-time if the OS is old) seems best for me. 



>>! In D6082#8, @foad wrote:
> OK, so we could go with Will's patch from D6044, or simplify even further to just:
> 
>   return (1ULL << 46) - 1;
> 
> What do you think, Will?

Roughly half of the systems I'm running on are old enough that they use the 44-bit address space, so I would prefer to allow those to continue working.  :-)     At least until they fail for some other difficult reason, at least. 
When the address space is wrongly calculated, the tests tend to hang or spin rather than crash and terminate,..  so if we decide that 44-bit address space will not be supported, we'll want to have an explicit abort of some sort.

http://reviews.llvm.org/D6082