Float to int conversion: Clamp the result to the integer range

[mbitsnbites]

Instead of always returning 0xffffffff for every out-of-range / overflow situation, instead do this:

	< min	> max	NaN
Signed	−2^31	2^31 - 1	0
Unsigned	0	2^32 - 1	0

Note: Several data processing systems (OpenCL, CUDA, SIMD, ...) prefer that NaN translates to zero.

[mbitsnbites]

According to: https://github.com/mbitsnbites/leanfloat