Clean up the convol and linmos'ing and fixed resolution mode

[tjgalvin]

If a desired fixed resolution is specified there are two rounds of convolution to a common resolution and linmosing that is performed in the self-calibration and imaging workflow. This is unnecessary. What should happen:

1 - convolve each beam image to the smallest common resolution
2 - co-add these optimal convolved images to a field image
3 - smooth the field image to the final fixed resolution image

This requires that the fixed resolution is larger than the optimal resolution, and there exists a kernel that can be used to do the smoothing.

Ultimately, this is intended to undo the original botched implementation where this silly pirate did not think.

[tjgalvin]

It would likely make a lot of sense during this to make a LinmosOptions to clean up the number of arguments that are being supplied to the linmos utility. Also may help to expose items to the cli.

[tjgalvin]

I have made some changes in this respect for #220. But I think there is still more to do.