Let XGA users be more flexible with entropy profiles

[DavidT3]

Currently the entropy profile class works in much the same way as the hydrostatic mass class - the user passes 3D temperature and density profiles, and models that they want to be fit to them, and then those models are used to generate the entropy profile.

This needs to change to give the user a bit more flexibility and better represent some of the different ways entropy profiles are calculated in literature. This will involve altering the SpecificEntropy class to allow for the following (in addition to the current method of passing 3D profiles and fitting models):

• Any combination of projected and de-projected temperature and density profiles will be allowed (though lets be honest it'll mostly be 3D density profiles and projected temperature - though I won't limit it to that).
• The user will be given a choice as to whether they wish to fit models and use them (this will no longer be the default however).
• If models are not to be used, then the user will be able to configure whether the data points are used as they are (which given the two profiles are very unlikely to have the same radial binning, will involve taking the one with smaller bins and finding the point in the larger-binned-profile that covers that region of space) or if interpolation is applied in some fashion.

[DavidT3]

For when the user selects data interpolation, I'm going to use linear/piecewise interpolation for now, though I may change it to 2nd or 3rd order polynomial interpolation (which I think should be sufficient for density/temperature profiles).

[DavidT3]

Was considering using the numpy interp function, but unfortunately it doesn't support being passed a 2D array of 'realisations' of the data (basically drawing different realisations of the data points using their uncertainty) and just interpolating along a specified axis. The scipy interp1d function does support this, but it is a legacy function and thus may well be removed in the relatively near future - I'm looking at the different 1D interpolation options scipy supports (https://docs.scipy.org/doc/scipy/tutorial/interpolate/1D.html#tutorial-interpolate-1dsection).

[DavidT3]

I think I really have to use linear interpolation at the moment, even though the scipy1d function is deprecated - if you look at the figures in issue #1168, the cubic splines really go a mad bit towards the outskirts with a projected temperature profile (helped by the larger uncertainties there I'm sure).

I acknowledge that its a quick fix using it here rather than figuring out something better, but I need to get this done.

[DavidT3]

At this point, interpolated data is now supported - here is a demonstration using 3D gas density and projected temperature:

The only thing left to add now is the data-point driven entropy measurement where the profile with fewer radial bins (generally going to be the temperature profile but I won't assume that) has data points matched to the points in the radial profile with more data points.

[DavidT3]

Data point matching now working if density profile has more bins than temperature profile (and now will be trivial to make work in the other circumstance):

[DavidT3]

I've thought of a failure case for the data-point version (without interpolation) - if the profile with more radial bins goes to a further-out radius then the we won't be able to find an appropriate value from the other profile. I think the simplest thing might be to do a check in the init for this circumstance, cut the radii off if necessary, and warn the user.

[DavidT3]

Think I've achieved everything I wanted to here, so the issue is closing. Next entropy-related adventure is adding models (see issue #1169 )