Fix bug in Math::BigFloat method bsin(). It can now handle any argument to
any precision. Also improve downgrading. This fixes CPAN RT #150590 and the
bsin() part of CPAN RT #107604.Fix bug in Math::BigFloat method bexp(). Specifying precision (rather than
accuracy) used to give a NaN. Also improve downgrading.Improve upgrading in Math::BigInt methods bsin() and bcos(). They now
upgrade only when necessary.