Stdlib: use float32 optimized trig functions in float32 module

[spotandjake]

Currently we just re-use the float64 implementations of the trig functions in the float32 library we should implement float32 optimized versions as they would be faster and the error would make more sense.

I am pretty sure we accumulate more error by using the float64 versions because Float32.pi is very different then Float64.pi