Pass SpectralAxis to physical models rather than energy edges

[jajmitchell]

This PR enables the physical models to be used with a SpectralAxis

By passing a SpectralAxis to the physical models, they are effectively evaluated at bin centers, but have access to the bin edges as these are stored as an attribute in the SpectralAxis class. This change will allow for a clearer path to using Astropy model classes for fitting as the dimensionality of the input and output will no longer be an issue.

[samaloney]

Nice is this ready for review?

[jajmitchell]

Nice is this ready for review?

@samaloney I need to just check that the fitting works with the physical models, then I think its good to go

[jajmitchell]

@samaloney the complications continue. The issue comes when fitting, specifically in Astropy fitting.py line 255, as can be seen below.

if isinstance(x, Quantity): 
    add_back_units = True
    xdata = x.value
else:
    xdata = np.asarray(x)

This means that for fitting, before being passed to evaluate, the fitter takes the value of the SpectralAxis object, which not only strips the units, but also all other attributes. I'm trying to think of ways around this, but haven't come up with anything yet.

[jajmitchell]

I have added a custom __call__ onto the ThermalEmission class, this is unnecessary for the ThickTarget as we evaluate at centers anyway