hash policy patch

[Paolo Carlini]

Hi,
> Hi
>
>     Here is an other patch proposal to refactor some code in hash 
> policy implementation
This bit I had yesterday in my local tree and made me nervous because 
__min_bkts is a float: std::lower_bound is templatized on the third 
argument, thus in principle the refactoring isn't really equivalent. Can 
we prove it is, given the actual values of __min_bkts?

Give me a second for the other hunks. In any case, if we'll come to 
agree about the substance of the issue, I think we should factor out the 
try / catch to a function, no? I see three identical instances of the 
same code...

Paolo.