Refactoring data loading functionality currently within Calc

[spencerahill]

C.f. discussion in #284, which I'll paste here:

Note that I do use Calc.load itself, still, so that should remain for now. Although I think it would still make sense to segregate it from Calc itself eventually...maybe a CalcDataLoader?

Yes I think there is a lot of room for improvement here. When plotting data produced by aospy it is often helpful to not just load individual calculations, but to load groups of them. For example something as simple as loading a particular calculation for a set of runs, and concatenating along a new dimension indexed by the run names can be very convenient. I've written various functions over the years to do variants of that where needed.

I have done the same. Most recently, I've generated a CalcSuite to dispatch the Calc objects I need for the data I want to load, and then loop through them calling their load method.

Within load, there are two pathways: from the primary output target containing netCDF files and from the secondary tar archive. So we need to account for both of them in any refactor.