refactor dsputils.rs to operate on generic floats

[itdaniher]

Hardcoding f32 / f64 really hurts the usability of a DSP library. It'd be a bit of work to refactor all the filter primitives to operate on generic floats, but not the end of the world, and probably a huge value add.

[itdaniher]

I'm not entirely sure this is doable, at this juncture. Constants suck.