Fixed Inf/NaN issues

Fixed some issues where some "infinity" and "Nan" values were propagating due to some calls to arithmetic to the effect of length(vec2(0.0, 0.0)) or normalize(vec2(0.0, 0.0)). This would particularly happen when the focal point lands exactly on a pixel position, causing there to be a black or white pixel right at the focal point! This issue has been fixed now