More muon-related fixes for the CP algorithms

First, this properly fixes the issue from !25203 (merged) (also reverted now in this MR). A try-catch block was added to the d0 significance calculation. As this will succeed in most of the cases (on bulk production I had very few affected events) this should not cause a large overhead. As having the track selection before the quality selection can be useful, I reverted the initial MR.

This MR also fixes the creation of shallow/deep containers. Note that the quality can also change depending on systematics, so I would also remove that somehow as with the pt cut, but I do not want to introduce another flag at this point so I might do it in a follow-up MR.