0.01 - first release

0.02 - fix scoping issue that prevented compilation with some (all ?) Microsoft compilers

0.03 - fix uniqnum for MS compilers

0.04 - provide more efficient version of uv_fits_double()

0.05 - In Uniqnum.xs, change the condition:
       #ifdef _MSC_VER
       to
       #if defined(_MSC_VER) && _MSC_VER < 1900

       I still don't know if "1900" is correct. I believe that the correct
       figure is greater than 1600 and less than or equal to 1924, but
       I haven't yet narrowed it down further.

0.06 - Further improvements to uniqnum().
     - Remove t/02old_uniqnum.t (no longer relevant).

0.07 - Alteration to the way we determine whether a UV or IV can be
       represented exactly by a double.
     - Fix handling of "DoubleDouble" NVs whose least significant double
       could be either 0.0 or -0.0

0.08 - On perls whose nvsize == ivsize, assign the bytes of the IV/UV to the hash keys,
       instead of the value. This is more efficient, despite the need to perform an
       additional sv_catpvn() of an extra byte.

0.09 - Move definition of NV_IS_DOUBLEDOUBLE from Makefile.PL to Uniqnum.xs.
     - No longer mess with the flags of the inputs.
     - Accommodate absence of SvUOK in perl-5.6

0.11 - Test, in t/01uniqnum.t, that uniqnum():
         1) does not alter the numeric flags of any of its numeric arguments;
         2) copies flags of numeric arguments to the corresponding returned values.