If an allocation is so huge that we cannot code the size needed in the failur...

If an allocation is so huge that we cannot code the size needed in the failure
object then we just return an out of memory failure object (instead of a retry
after GC failure object).  Not all places that checked for retry-after-GC were
able to handle an immediate out of memory failure.

This fixes http://code.google.com/p/v8/issues/detail?id=70

Committed: http://code.google.com/p/v8/source/detail?r=477