Simplify overloading of NVs by utilizing Rmpz_set_NV(). This also seems to fix the off-by-one errors that might afflict NV overloading when NV type is doubledouble.
Limit test in Makefile.PL for GMPQ_PV_NV_BUG to perl-5.35.9 and earlier.
In Makefile.PL, insert '-lquadmath' link in LIBS entry in %WriteMakefile - but only for Cygwin -Dusequadmath perls.
Workaround Cygwin -Duselongdouble printf "%.0Lf" formatting bug. For example of bug see: http://www.cpantesters.org/cpan/report/981d0446-6f3a-1014-84d5-f4f801469742