Update of calibrate measurement size should become internal to the MTJ backend

[tboldagh]

In the MTJ implementation we have such line:

acts/Core/include/Acts/EventData/MultiTrajectory.hpp

Line 931 in 5cab50d

calibratedSize() = measdim;

This sets the size of the calibrated measurement. However with different backend (xAOD in ATLAS) we do not need this extra size as the data is kept in vector<vector> and the size is known. We can add a fictional method to update the size but
it seems that we can eliminate this method:

acts/Core/include/Acts/EventData/MultiTrajectory.hpp

Line 888 in 5cab50d

template <bool RO = ReadOnly, typename = std::enable_if_t<!RO>>

and let the allocateCallibrated to handle this update.

Tagging @paulgessinger
and for info @CarloVarni Marcin Wolter

[paulgessinger]

So you would make calibratedSize to be purely read-only?

[tboldagh]

So you would make calibratedSize to be purely read-only?

Yes. And the update of the size handled by allocateCallibrated which would be backend dependent anyways.

[tboldagh]

That was quick. Thanks. There may be another small issue here. The calibratedSize is now using "component" to get to the actual size. In ATLAS xAOD implementation this would require that we somehow get the pointer to the vector size property. Not sure if that is doable. Not sure how to best solve it. Duplicating this information in the vector & in mesdim variable seems like asking for trouble.

[paulgessinger]

Oh that's a good point I didn't consider yet. I suppose we could instead special case the calibratedSize call to either call a fixed function, or change the component function to return that by value and not as a pointer.

[tboldagh]

We can indeed specialise it! Nonissue then.

[paulgessinger]

No I believe you're right than in the current implementation it tries to cast it to a pointer, which will break if the backend passes the size by value. So I think some form of fix is still needed.