Polynomial baseline fitting

[MatthewWhiting]

A further baseline fitting mode that would be good to implement is polynomial fitting. This should also have iterative sigma-clipping (to avoid the bright sources that will inevitably be present).

The iterative sigma-clipping should iterate until the total change is less than some value, after rejecting points in the residual spectrum that are outside n-sigma.

An alternative approach could be to fit only to a given set of channels. We could use the same interface as for the flaggedChannels parameter. This would only work if a user knew that there were specific channel ranges that did not have sources (or interference for that matter) in them and yet still gave a good representation of the baseline.