#55 Beam fitting incorrect for forced-circular beams

[André Offringa]

For beams smaller than one pixel, the Gaussian fitter sets the min/maj axes to one pixel to overcome very elongated beams or fitting instability. This does not preserve the integration, so is not correct. Best way is probably to do a re-fit keeping one axis set at one pixel and having the other one fitted.

Sort of the same holds for circularly forced beams, in which the maj/min or changed after fitting.

[André Offringa]

Circular fits are now corrected, will open a separate bug for small beams.