v0.1.3

This release add several new functions listed below, as well as a bug fix for powi when using large negative exponents. Fuzzing of all of the listed functions against their references in libm/compiler_builtins is also now implemented and passing. Fuzzing is exhaustive over f32 values and an equivalent number of f64 values spread over the possible value range.

Also, thanks to @elpiel for fixing the nightly-only doctests and adding Default/From implementations for both types.

New SoftF32/SoftF64 methods:

• sin
• cos
• floor
• round
• trunc
• copysign