[m-rev.] for post-commit review: speed up string.format for integers some more

[Julien Fischer]

On Wed, 12 Nov 2014, Zoltan Somogyi wrote:

> This diff speeds up formatting integers by another 12% or so,
> taking the time on that benchmark down to
>
> 15.841u 0.007s 0:15.86 99.8%	0+0k 0+0io 0pf+0w
> 16.051u 0.004s 0:16.06 99.9%	0+0k 0+0io 0pf+0w
>
> Does anyone have an objection to raising the limit mentioned
> in the diff from 2^15-1 to 2^31-1, i.e. to assuming that
> Mercury platforms have at least 32 bit integers?

No, go ahead.  That assumption is pretty safe.

Cheers,
Julien.