gmath v0.2.0

Improve usability and fix bugs. Remove all Go module dependencies.

Significant changes:

• Remove all Go module dependencies.
• Copy type constraints from the golang.org/x/exp/constraints package.
• Prevent all NaN float64 -> float32 conversion, which converts "signaling NaN" to "quiet NaN".
• Only accept floats in Logb and Ilogb
• Remove generic Inf and NaN, which require awkward syntax to use.
• Add Inf32 and NaN32 for generating float32 infinity and NaN values.
• Assert on binary equivalence of float values in tests to prevent logical float equivalence from hiding errors (e.g. 0 vs -0).
• Add test cases for special behaviors.