Irregular behaviour of polyhedralprismcurrentsource cross section when used with TF PictureFrame circuits

[kj5248]

Describe the bug

When attempting to use ArbitraryPlanarPolyhedralXSCircuit for TF coils noticed weird behaviour of the winding pack position, however this didn't appear for other parameterisations such as PrincetonD. Tracked this problem through to the polyhedralprismcurrentsource class, specifically the _calculate_points function. As far as I could tell the issue was related to using _local_to_global on the cross section when the cross section coordinates were already located around the global center for the current source. Not sure why this only occurs sometimes, assumed it got missed in the tests due to only using cross sections which were around the origin.

First image below shows the intended prism, obtained by setting the XS coordinates to be centered around 0 whilst the second image is if the coordinates are centered around the shape center.

This could be solved by enforcing that the XS upon initialisation is in local coordinates or if not makes it so that it is. This could be done within the set cross section function.