- Third attempt at getting the assignPV.t workarounds right

 - Try again to get the assignPV.t workarounds right

 - Workarounds in _atodecimal() and assignPV.t to cater for broken
   compiler/libc handling of signed zero _Decimal64 values.

 - Add PV overloading.
 - Rewrite assignPV in XS. (Now uses _atodecimal XSub.)
 - Rewrite PVtoD64. (Now uses _atodecimal XSub.)
 - Remove (unused) XSubs _c2ld() and _c2d().
 - add assignD64() and use it to modify sub assignME() such that the latter should always
   work correctly even when $Config{longdblsize} is 8.

 - Change D64toME to use decode_d64() rather than C's strtold (as per Math::Decimal128).
 - Fix bug in overload_sub by adding overload_neg.
 - No longer allow new() to assign an NV. (NVtoD64 must now be used instead.)
 - Add assignIV(), assignUV() and assignNV().
 - Add D64toFSTR() and D64toRSTR().
 - Fix Decimal64.pm's FR64toME to work with Math-MPFR-3.23 and later (when
   Rmpfr_set_decimal64 was renamed to Rmpfr_set_DECIMAL64).

 - Add get_exp() and get_sign()
 - Add d64_bytes()

 - Add unpack() and associated subs.

 - Fix t/conv_math_mpfr.t to cater for the case that Math::MPFR loads but is an older
   version that doesn't provide the Math::MPFR::_MPFR_WANT_DECIMAL_FLOATS function.
 - Add metadata to Makefile.PL.
 - Work around compiler/libc bug in handling of the value 897e-292. (Affects PPC only.)

 - Fix D64toME() so that it works correctly on powerpc processors.
 - Add FR64toME and pFR.
 - Document LDtoD64 and D64toLD.
 - Add assignInf, assignNaN, assignPV.
 - Allow overloaded operations to accept integers (IV/UV).
 - Alter the way that strEQ gets hold of HvNAME(SvSTASH(SvRV(sv))) - so that -Wall
    doesn't emit "null argument" warnings. (Thanks Daniel Kahn Gillmor.)
 - Have only *one* NaN ( = inf/inf). Originally we could also set NaN to -(inf/inf),
   but this is stupid and can cause problems.

 - Fix bugs in t/lengthy_mantissa.t. (A '-' in a regex needed to be escaped.)
   Also the precise number that appears in $@ for test 3 will vary, depending
   upon nvsize. (Test 3 modified to accommodate this.)

 - Add LDtoD64 and D64toLD (conversions between Math::LongDouble and
   Math::Decimal64 objects).

 - Small change to _sci2me() to avoid 'uninitialized' warnings on 5.10 (and
   perhaps earlier).

   Replace 'exp10' (which already exists in at least some C libraries)
   with Exp10.

 - First release.