Properly account for pixel resolution (versus pixel size)

Currently we use the pixel size in the covariance matrix we pass to ACTS. This should really be 1/sqrt(12) times the pixel size, as this is the sigma for a uniform hit distribution in a square pixel.