Correctly subtract the dark signal seen on r3

There is a bright defect in the r3 camera which doesn't subtract out well with the dark frame processing, presumably because its amplitude varies with time.

Please change the code to estimate the amplitude of the defect based on the part of the CCD where there are no fibre traces.

Note that this cannot, in theory, simply be done by adjusting the amplitude of the dark frame as there may well be a stable and unstable component. An interesting experiment would be to carry out a PCA decomposition of the darks and measure the dimensionality of the needed solution.