1.07
     - Fix reference counting of $Math::Ryu::pinf/ninf/nanv in fmtpy() XSub.
     - Add more fmtpy_pp() tests.
1.06
     - Wrap and export SvPOK (as ryu_SvPOK) and SvIOKp (as ryu_SvIOKp).
     - Optionally allow (via setting of $Math::Ryu::PERL_INFNAN) infs and nans to be
       stringified by nv2s() to the same as perl provides. Previously they have been
       stringified by nv2s() strictly as "inf", "-inf" and "nan" - which could differ
       from perl's presentation. See "DISPLAYING INFS AND NANS" in the documentation.

1.05
     - In 1.04, I failed to amend pany() and sany() to allow for the "PV_NV_BUG". This
       is now fixed.
     - Add is_NV() function which identifies whether its argument is a perl scalar
       floating point (NV) value.

1.04
     - Reinstate detection of the "PV_NV_BUG" in older perls, and fix a bug in t/sprintf.t.
       Fixes (eg) https://www.cpantesters.org/report/72adab5e-0567-11ef-98b0-b3c3213a625c

1.03
     - Add new "print" functions pany() and sany(), and "sprintf" function spanyf().
     - Export ryu_SvIOK and and ryu_SvNOK - formerly known as the unexported
       functions _SvIOK and _SvNOK.
     - Remove the "PV_NV_BUG" check as it is irrelevant. (Subsequently re-instated in
       version 1.04 because, as it turns out, it wasn't irrelevant.)

1.02
     - Fix fmtpy_pp() to handle values like '6E15' correctly. (In 1.01, they appear
       as '6e+15', but python3 has '6000000000000000.0'.)
     - Rewrite fmtpy() in C. (The earlier pure perl implementation is still available
       as fmtpy_pp.)

1.01
     - Fix s2d() to handle Inf and NaN.
     - Add new functions pn, pnv, sn, snv.
     - Revise fmtpy() to be more efficient and, hopefully, more readable, too.
     - Conditionaly add -std=c99 flag to compilation of libryu - aimed at
       avoiding some "UNKNOWN" reports and allowing the module to build on those
       affected platforms.
     - Add benchmark script (benchmark.p) in "benchmark" folder.

1.0
     - Significant changes between 1.0 and 0.7 include removal of some fairly
       useless functions (which can be reinstated, if requested), and support
       for -Duselongdouble and -Dusequadmath builds of perl.
       A debug version of this module is no longer provided - also reinstateable
       on request.

0.07
     - Workaround bugs in perls older than 5.30.0.
       (Math-Ryu-0.06 failed to allow for these bugs in some of the test sripts
       and I don't know if 0.07 covers them entirely ... time will tell.)

0.06 - Extra tests added to t/mpfr_checks.t and n2s.t.
     - Add new function fmtpy() - and also add t/reformat.t to test it.
     - Silence pointless compilation warnings generated by the XS files

0.05 - Add n2s function - same as d2s except that n2s also does what you mean
       when its given argument is an integer (IV/UV) or a string (PV).

0.04
     - Add _has_uint128() function (in both d2fixed.c files and both .xs files)
     - Patch s2d.c and Debug/s2d.c so that 32-bit Visual Studio (which does not
       provide the _BitScanReverse64 function) can build Math::Ryu.

0.03
     - Expand coverage to include environments that don't provide __uint128_t
     - Add Math::Ryu::Debug.
     - Add s2d().
     - Ensure that Newxz() is defined.

0.02
     - Fix big in t/mpfr_checks.t (No allowance was being made for the possibility
       that atonv() or nvtoa() might not be available.

0.01 - First release.