Optimize hashtable allocator

[Paolo Carlini]

On 11/02/2012 11:34 AM, Jonathan Wakely wrote:
> As Francois pointed out, the printers for tr1 hash tables don't work 
> any more, and I'm still seeing PR 55041 and confused by it (and also 
> confused by F16's gdb being unusable with GCC trunk). Either I'm being 
> dumb and failing to see something or we need Tom to help out with 55041. 
I see. To be safe, eventually better adding separate tests for the tr1 
variant of things: in the future certainly will not track the C++11 
features.

Paolo.