Allow numpy args for conversion functions

[dkirkby]

Some functions already work with numpy args (e.g., xyz_from_rgb) but others use functions or constructs that are not numpy aware (e.g., rgb_from_irgb uses float() and lab_from_xyz uses scalar '>'). This issue is a feature request to make all functions numpy compatible so that arrays can be converted with a single call instead of requiring (much slower) loops.