The implementation of the expt_d functionality is now implemented in the
mp_expt_d_ex() function.
The user can now choose between the old (more timing resistant) version
and the new version by modification of the parameter 'fast'.
mp_expt_d() defaults to the old version
The loop was always iterating DIGIT_BIT times, instead of
halting when possible. This changes makes sure it executes
less operations. This change has also been incorporated
into Rubinius / https://github.com/evanphx/rubinius which
uses libtommath
