0.6.0

This release contains some breaking changes which should be easy/trivial to accommodate.

Fixed:

• Fix unsafe code in NonZero conversion on input values which wrap to zero (#52)
• Fix: allow overflow of large u128 values to f32::INFINITY (#51)

Changed:

• Adjust ConvApprox<f64> for f32 to behave more like *_f64 as f32: too-large values round to infinity (instead of a range error) and rounding uses roundTiesToEven. Unlike as f32, NAN still yields Err(Error::Range). (#47)
• Change Conv<f32> for f64 to trap NAN (#48)

Removed:

• Remove support for Saturating, Wrapping (#49)
• Remove assert_nonzero feature flag (#52)

Full Changelog: 0.5.5...0.6.0