Refactor the TF winding pack geometry to prevent negative values

[mkovari]

The problem of negative values appearing in the geometry of the supercon TF coil has been raised several times:
#515, #785, #1709, #1767. (I will close all these.) It has never been properly solved.

The best way to prevent negative numbers is to avoid subtraction. The winding pack would start from the inside (cable diameter), and be built up from there.

The radial build of the TF coil can probably be sorted using the method in #515, and the radial build constraint will ensure everything is consistent.

The toroidal build doesn't have a constraint at present. Either we will have to add a constraint or be clever enough not to need one.

Note that there are several different options for building the supercon TF coil.

@jmorris-uk @ajpearcey @jonmaddock