Fix 2D local to global methods for ITk Strips

When porting some code to master I noticed something strange when I was comparing back to 21.9 using ReadSiDetectorElements - some local-to-global tests were giving sensible results in master but in 21.9 they looked incorrect, as was discussed in an meeting with @asalzbur @tadej @ncalace and @tstreble

I found why, and this is because in 21.9 some inline methods were using the origin instead of the centre for the reference point of the local coordinates. For pixel and strip-barrel modules (and all Run 2) these are the same, but for ITk strip endcaps they differ due to the annulus shape always having a frame with an origin on the beamline.

The results in 21.9 are definitively wrong as a result, but I don't know for sure if perhaps this was accounted for somehow "downstream" or we just never use these particular versions anywhere important. Since these are very generically named and also inlined, its tricky to tell where they could be being used (and also tricky to test the impact locally).

I think this is something to be fixed, and hopefully CI test (and worst case ART) will show if there are any unexpected effects because this was "fixed" downstream somewhere.