Patch to update top-level config to autoconf 2.59

[Paolo Bonzini]

> The problem is that it doesn't necessarily have to be gas - it could be
> the native assembler.  GCC has:
> 
>   m4_pattern_allow([AS_FOR_TARGET])dnl
> 
> So it's probably better (and a smaller patch) to just do that.

Yep.  Also, it would be best if you could test this change also on a 
cross-compiler, a Canadian cross, and maybe even a host-x-host (i.e. 
host != build, but build == target) configuration.  Be sure to use 
non-canonical names (i.e. names that config.sub will change).  :-)

Personally I would not be against this change even if it breaks some 
configurations, as long as it is known *which*.

Paolo