Doc bug found and suggestions

[epl]

   The definition of  mpf_get_d_2exp() in the manual is incorrect. A
(signed long int *) is the type of the first argument rather than
(signed long int). I've found this mistake in both the postscript
manual and the html version.

And two suggestions
- A method of determining logs would be nice.
- Are there any suggested methods to find out if a mpf_t type has
already been initialised? One method I'm thinking of using is to
manually initialise the mp_limb_t pointer to NULL and reset it to NULL
on clearing. However, I'm not sure if it is acceptable as it delves
somewhat into the internal structures of GMP.

Eddie